LT3650 Lion battery charger heat issue

[Moonwalker]

Have someone ever used the LT3650?

I have designed a circuit charging a 18650 battery. It works ok but the IC gets very hot (50C) at 12V input and over 65C at 28V. The IC has about 20mm² of copper on the bottom side and about 10mm² on the top side. Charging current is set to 1.5A. IC should have thermal shutdown but is this heating normal?

I build about 10 units and all have the same heat issue but now got 2 back with blown IC. Client said he plugged them in cigarette lighter socket of a truck but that should be 12V and if by mistake it was 24V it should still not get over 30V while the IC takes a max of 32V.

Any help?

 

[treez]

i would check no shorting between pads. -they are very close together.
Also check you have used shottky diodes for the bootstrao and buck diode.

 

[Moonwalker]

Sure I checked for shorts and diodes are shottkey type.
IC has an efficiency of 84% at 12V and 82% at 20V so power loss should be around 0.5W at 12V input.

 

[treez]

http://cds.linear.com/docs/en/datasheet/36504142fc.pdf

lt3650 has 1mhz frequency so the switching losses may be significant even tho’ vdc in is low.
But anyway, have you got good thermal coupling of the heat pad to the pcb copper?…with multiple thermal vias to the bottom copper and intermediary layers….ie, plenty of cooling copper on the PCB?

- - - Updated - - -

Also, which is hotter, the schottkey or the IC?
The internal fet of lt3650 has a rdson of 0.175 ohms, so double that for high temps, ie, 0.35 ohms, so it definetely will need decent cooling copper with 1.5A output current.

- - - Updated - - -

Right, in the lt3650 datasheet it states that the boost switch drive is “30mA/A”.
So 30mA is actually a lot of drive current, as I notice the internal switch is not a fet….so when your vin is 28v, then that 30mA is going through the 28v, and this could produce quite a bit of dissipation in itself…….at vin = 12v, your switch duty is about 45%, so you have a dissipation due to drive current alone of 0.45*12*0.03 = 162mW , which isn’t insignificant for such a tiny tiny chip as the lt3650

…I mean look at the size of it…titchy little 4mm by 3mm….think of an equaivalent size smd resistor, ie the lt3650 is about the same size as two 0805 resistors, and that is 300mW abs max (for the two of them)…..so you can see this is a problem for your chip because IC’s are nt made of as good thermally condicting stuff as smd resistors. So plenty of thermal copper

 

[mtwieg]

IC has an efficiency of 84% at 12V and 82% at 20V so power loss should be around 0.5W at 12V input.

Did you actually measure efficiency or are you just reading off the datasheet?

 

[Moonwalker]

@ mtwig - No did not measure efficiency - just did a quick calculation taking reading from datasheet

@treez - Only have 2 layers - (first post copper data is incorrect) top layer has about 150mm² copper and bottom layer about 400mm² with plenty of filled vias connecting them- maybe thats not enough copper for that power?
- Schottkey is a 5A. Its the IC that is heating up. Temp on IC and on backside copper is similar so the thermal coupling seems to be working. I cant find any reference to RDSon on the datasheet.
- Yes it very tiny but it is designed for this job and it should run 2A while I am running at 1.5A.

 

[treez]

i suppose all you can do then is add more copper, or perhaps stick a bit of aluminium on the bottom thermal copper layer..via an insulation pad
And indeed, do an efficiency test to see if there is more dissipation than expected.
I trust that you have actually treated the heat pad as a pad and not accidentally covered it on solder resist?

 

[Moonwalker]

Umm no it is covered in solder resist actually. Would that make a big difference?

 

[treez]

yes, you need to have no solder resist on that pad, it has to solder to the underside of the chip for thermal reasons

 

[Moonwalker]

No I mean the other side. Under the IC it is solderable ofcourse.
What I was asking is if the copper pour would radiate more heat if it is bare (silver plated) vs covered with solder resist?

 

[treez]

will convect more heat away if not covered with solder resist.
However, if you end up putting a heatsink on it, via a thermal/insulating pad, then you might as well leave off the solder resist.
By the way, did you do the efficiency test?
That way we can see if there is an abnormally high dissipation going on in there.

- - - Updated - - -

Does it still get hot even when running on light load?

- - - Updated - - -

What are your circuit waveforms like, eg switching node and inductor current, do they show normal behaviour?

- - - Updated - - -

Inductor is running hot,? maybe it is transferring its heat through the pcb copper to the ltc3650 copper?

- - - Updated - - -

It is automotive use?....maybe you have input LC filter, and when the relay switches it on the LC ringing goes over the 40v abs max of the ltc3650 chip?
Be wary of that.
Can you do thermal test on it in chamber?
Ie, is it only failing when hooked up to the lorry battery in the lorry, or does it fail also when connected to a 28V PSU in the lab or thermal chamber?

 

[mtwieg]

@ mtwig - No did not measure efficiency - just did a quick calculation taking reading from datasheet

Then you should measure the actual efficiency and power dissipation. The datasheet numbers are generally optimistic and can't be taken for granted.

 

[Moonwalker]

Vin to IC = 11.79V
I in = 0.578A

Vout to battery = 4.199V
Vout before Rsense = 4.289V
Rsense = 0.068R --> Iout = 1.324A


Power In = 6.815W
Power Out = 5.559W

Efficiency = 81.6%

Power loss = 1.138W

Now this power loss makes more sense for the heat being generated.
Here is a picture of the voltage at the coil





After battery is almost full and I out drops to 0.735A -- Pin = 3.454W - Pout = 3.116W - Eff = 90% - Ploss = 0.34W and IC temp is much less.

 

[mtwieg]

For some reason the thermal impedances for the packaging is not given in the datasheet, but can be found here. For the DFN package the TCA is around 38C/W and for the MSE package it's 25-30C/W. So your temperatures sound reasonable. 65C feels hot, but the IC should be perfectly capable of withstanding it, so long as the PCB temperature is kept low. Or is this thing kept in an enclosure with restricted airflow...? Are you certain the failures are due to the IC overheating? Automotive environments are notoriously harsh, and electronics will need additional protections (load dump, EMI filters, etc).

 

[treez]

Well, as you can see in the attached excel, your measured loss is sounding what it calculates to be…assuming a vf=0.4v schottky, and an 4.7uh inductor with 40mR of esr.
I also assumed a 10ns fet switching transition time. I also assumed 1.3a of output current, since that’s what youd get with a 0.062 ohm sense res and a 4.7uh inductor.

Just to check though, what inductor and diode are you using (part numbers)?

If you use a too-high-current-rated schottky, then this might be all very well for the schottky, but it does mean higher switching losses for the fet in the lt3650.

I calculated about 413mW of loss in the lt3650 at 12vin, but this assumed 10ns fet transition time, and a schottky with very low capacitance.

So yes, what thickness of copper do you have on the pcb. Also, when the guy plugs it into the ciggy lighter, is there any damaging overvoltage ringing on the input bus…..this is a common killer. If you have an input LC filter ahead of the LT3650 then you need to put a diode (cathode to input) across the inductor otherwise your input filter will ring like mad when vin is suddenly applied and it will wreck your lt3650.

You might also like to put thermal copper pours on all the inner layers under the chip too.

Also, of course, are you sure the customer hasnt connected something like a 28v battery to the output and then plugged in a 12v input………all these kind of abuses might have been subjected to your cct.

Without a proper suite if thermal tests being carried out in a thermal chamber with the cct in its intended enclosure you will never know what was the problem.

 

[Moonwalker]

@mtweig - Have no idea how the failures are happening but the temperature was my only concern so far. The board is kept in a closed enclosure however the IC should have thermal protection and should reduce output current if the temperature goes too high.

@treez - **broken link removed**
https://uk.farnell.com/multicomp/mcsch664-100ku/inductor-10-h-10-radial-leaded/dp/1864362
From the scope picture I posted the risetime/falltime seems longer than 10ns - more near 50ns. Input Diode only drops 0.27V at 0.58A.

PCB copper should be standard 50oz.
Have no idea yet what is happening at the client - if I get another returned unit I will have to go do some testing there. In the lab I had no failures at up to 28V but I am listing the unit for 12V operation only.
This is simple PCB with 2 layers only.
The unit only has access for a DC socket for charging with reverse polarity protection so clients cant access battery.
Dont have access to thermal chamber or camera. Only have an IR thermometer.


Now can you guide me to a similar Lion Battery charger IC taking 12V input and charging at 1-2A with better efficiency?

 

[treez]

Thanks, I am a bit confused when you say there is no access to the battery, isn’t it a battery charger? Is the battery fixed in to something then?
Anyway, the diode is 5A rated and has a fairly significant capacitance, so that would increase switching losses in the lt3650, so you could look for say a 3A schottky with lower capacitance.

Automotive load dump is something to be wary of for anything on the automotive bus……that can mean voltage going up to 70V or so. You might have to put a series fet switch in ther so that it switches your unit out when the vin goes above 30v….also add a tvs downstream of it because the turn-off action will obviously not be immediate.

I don’t think you can get 50 oz copper, its usually 0.5,1 or 2 oz

If the transistor rise and fall times are 50ns then that’s a lot worse for dissipation, I calculate 1W of loss in the lt3650, and that doesn’t include the high capacitance of the schottky.

So does you charger just keep slamming 1.3A into the lithium battery, with no overvoltage checking of the battery? Maybe the battery went short and your lt3650 didn’t like that. Have you tried shorting the load to test what happens.? The current will staircase up high.
But also, we don’t know if its ingress of damp into the enclosure, maybe its getting left in the truck cab all night.

Regarding alternatives, I think there is lots of ones that have the fet external and then you can be sure to have less dissipation in the controller. That might be best if its enclosed in a hot lorry cab which literally gets above 70degC ambient with the sun shining through the window when the cab is parked up….if it was left charging your battery in that condition then that wouldn’t be good for the lt3650

- - - Updated - - -

maybe you can do one with an external fet nxt time, and just deluge it in conformal coating so that it doesnt get damp problems......as you know, damp proof enclosures usually relies on the assembly staff remembering to smear silicone gel over the enclosure mating sections before mating it, and they may forget, and then its death to your lt3650 etc

- - - Updated - - -

All the monolithic controllers have high rds on fets in them....you should use external transistor for your enclosed setup.

 

[bill66656]

treez's analysis and calculation were impressive, and standing on his shoulder I'd like to give below advice.

1.5A was fairly below the LT3650 max. current limit, so it won't blow as long as the case temerature wasn't very high such as your experiment case. You'd better consider how to lead the heat from the exposed pad to the case surface. Also following the thermal design guidance in the datasheet to extend the exposed pad on the PCB as large as possible.

Still can add a warning notice to inhibit to use the production as the vehicle room temperature rises above 45C or some degree, synchronously you need to ensure it can stand that environment temperature.

Finally a remind, the 12V battery voltage of the vehicle will be about 14V in reading.

 

[Easy peasy]

50 - 65 deg C is not that hot, obviously you want to avoid thermal run-away and resultant shut down when it is cased up and in a high ambient, so adding some SMT heatsinks (even QC tabs) will be a big help, although for a fully cased (sealed) box the ambient will rise, limited only by the thermal impedance thru the walls of the box and the amount of convection inside it - often this is a very limiting factor, if there are no holes for air flow.

 

[treez]

So yes, if you do suspect temperature is the problem, then I would put in a cheap 6 pin PIC micro and get it to read the output of a temperature sensor inside the enclosure…..

Temperature sensor
**broken link removed**

This is what I believe we should have done when I worked in automobile lighting. –Because when you have something inside an enclosure that is in the sunlight, its internal temperature gets ridiculously high. –And your charger is in an totally sealed enclosure inside an automobile cab that could be sat there in bright sunlight.

….So , you could get the charger to switch off if the temperature goes above x degrees…..don’t rely on the temp sensor inside the LTC3650, these are always “ last resort” and I notice that safety agencies won’t respect them.

Alternatively, if you don’t want the charger to switch off, then what you could do is get the micro to blow a fuse inside the unit so as to indicate that it went above temperature “x”. That way, when you get a failed unit back you can see if the unit suffered overtemperature before it packed up or not.
The fuse would just be say a 200mA fuse that doesn’t actually lead anywhere, but when your micro turns a transistor on which makes current go through this fuse, it blows…..and then you have an indication of whether or not an overtemperature event actually happened when you get a returned failed unit.

Because at the moment, we don’t know if your problem is….

1…ingress of damp
2…overtemperature
3…Overvoltage
4….Conection to short circuit output
5…dry joint on the buck diode, say
6….dry joint of the heat pad of the lt3650
7….wrong inductor fitted
8…buck diode the wrong way round.
9…etc etc

By the way, what is doing your charge control algorithm?
Also, when you get a failed unit back and you power it up, what happens?