When to heat sink or not?

[BrendanP]

Im using the MAX712 to charge 3X 3500 mAh NiMH's. On fast charge I charge at .875 mAh's. Im using a 12V supply to the charger circuit. I use 12V because I want to be able to charge the device off a car eventually.

As per the Maxim data sheet 'typical opperating circuit' Im using the a 2N6109 NPN transistor to controll the current flow into the batteries. pdf below.
http://www.onsemi.com/pub_link/Collateral/2N6107-D.PDF

I notice that the 2N6109 becomes very hot when charging. At the moment I have the pcb I designed running on the bench, I use a bulldog clamp as a make do heat sink. I have to fit the board into a tight enclosure so it would be easier not to have to use a heatsink.

How do I decide if I need a heatsink or not? Ive looked through the data sheet and cant see any obvious guide. I see there is refference to the devices power dissipation ability and opp. temps. Is there a way to work it out by formula? Can I simply leave it to run on the bench for a few days and if it keep going conclude it doesn't need a heat sink?

 

[BeanieBots]

All the data you need is in that spec sheet.
The theory is very simple but the practice can be a little tricky.
A component will have a maximum power spec and a maximum temperature spec. Heat transfer (at sub glowing red) is linear and is expresed as degrees per watt. That is, how much the temperature will rise for a given power.
More often than not, the maximum power spec is quoted assuming the package case is at 25C. ie, it has an INFINITE heatsink attached. Hence the limiting factor is the thermal transfer from junction to case.
For example, the 2N3055 transistor is quoted as 90W. With no heatsink it would melt with only about 10W.
So, if you know the C/W for the bare package, you know how hot it will get for XW of power. That's degrees above ambient. In an enclosure, ambient will rise so you need to take that into account.

If you are going to make your own heatsink, have a look at some readymade heatsinks to get a 'feel' for the C/W value of a similar size.

 

[Dippy]

I'll go along with BB there.
You should be able to calculate approx the power dissipated in the transistor. You can measure worst-case current and you can measure the Vdrop across C/E. The data sheet will give figures, but there's nothing like watching it as the device gets hot.

Once you've got your 'worst case' watts then look at BB's heatsink calcs.

But in answer to the question "How do I decide if I need a heatsink or not?"
- well, my Dear Watson, if it gets F-Hot then you need a heatsink. (This assumes your circuit and component choice is optimal of course!!).

Another thing: you said "2N6109 NPN", well which is it? Have you accidentally fitted an NPN or have you used the 6109 - a PNP.

"Can I simply leave it to run on the bench for a few days and if it keep going conclude it doesn't need a heat sink? "
- if it gets very hot very quickly as you have described then this would be a very silly, sorry, unwise thing to do. Semiconductor lifetimes vs temperature? Check it out.

I don't know the current you're at with your design, but I'm a little surprised it's getting to frying pan temperatures. Are you sure you've connected everything up correctly.

 

[BrendanP]

Sorry Dippy it is a PNP. I posted at about 1 am here and misread the data sheet.

It gets hot to the extent I can't touch it, I have noticed that if I use a lower voltage input it does not get as hot. I have a rule of thumb if a device gets that hot then it needs heatsinking but I didnt know if there was any real basis for that idea.

The circuit goes into trickle charge and the current drops back as it should after V across the battery levels out or timer runs out. I presume from this all is as it should be.

Thanks for the replies guys, I'll carry out both of your suggestions.

 

[BeanieBots]

As a rough guide only:-
Max. case temp < 70C (or junction < 150C whichever is the lower)
Too hot to hold ~50C.
T0-220 "naked" in free air ~50 C/W. (ie ~ 1W max.)

 

[boriz]

I am about to build a PICAXE controlled charger for my electric RC plane. NiMH 8 cell, 9.6v, 1000mAh. The plan is to use a 13.8v PSU with a cigar-lighter socket, and control the charge rate using PWM instead of a linear current regulator. PWM is much more efficient. Since the transistor is always fully off or fully on, the heat dissipation is tiny. No heatsink required.

I plan to use 2 * PICAXE-14M to charge 4 packs at the same time. Each 14M will monitor the temperature of two packs. The charge is complete when the temperature rises 20C above start temperature, or the rate-of-temperature-rise gets over 1C per minute.

I will use a modified pulse charge method. The MOSFET turns on and off at several hundred Hz, around 20% duty, supplying 13.8v through a 0.5R power resistor. The exact duty will be determined by experiment to get an average charge current of 1 Amp. This continues for one second. Then the MOSFET is left off for another second. Then repeat. Constantly monitoring the temperature. The total average charge rate will be around 500mA, charging the packs in about 3 hours.

Parts are on order

BTW, a good rule of thumb is that if you cannot hold your finger continuously on the component, it’s too hot and needs a heat sink.

 

[BeanieBots]

Or a buck SMPS if you want smooth constant current.
One FET, one inductor, 2 diodes and a few caps.

 

[westaust55]

Brendan,

You say . . .

I'm using the MAX712 to charge 3X 3500 mAh NiMH's. On fast charge I charge at .875 mAh's. Im using a 12V supply to the charger circuit. I use 12V because I want to be able to charge the device off a car eventually.

So the data is 12Vdc input and 3 x 1.25 or allowing for a slightly higher voltage to get a charge into the battery (assuming 3 in series) then 4 Volts across the battery.

You mention FAST charge at 0.875mAhs.
[Amps is current flow, Ah is energy capacity]
Do you mean 0.875 Amps (875mA)?

Assuming so, then the voltage across the transistor is 12-4 = 8 Volts.
(More if the batteries are in parallel - but that's not the best and another story)
If the current is 0.875 Amps so the transistor must dissipate 8V x 0.875A = 7 Watts

That’s a lot of heat and a heat sink is a definite necessity.

With the TO-220 type transistor case the tab at the top is connected to the collector leg of the transistor so as well as a heat sink you should include an thermally conductive (mica) insulating tab. Irrespective of the insulator, some heat sink compound is also required.

 

[BrendanP]

Once again westaus another typo due to posting late at night. I meant 875mA.

I thought of using the 28X1 that controll the GSM module to handle the battery charging as well. I just dont have the memory space to spare.

Given the high cost of the MAX712 (around $6) a picaxe controlled charger might be the go.
I could use seperate picaxe that wouldnt take up any more space than the MAX712 to do the same job and give more functionality as well.

I thought of measuring the V rise across the battery picaxe adc and terminate charge when it reaches the right levels, I could add temp monitoring as well...... refuse to fast charge, trickle only if batteries too cold etc...

Boriz what do you use to monitor temp?

Thanks for the thought provoking tips guys.

Yep, I've given it more thought and I'm going to give the MAX712 the sack and use a dedicated picaxe to do the job.

 

[moxhamj]

Worth getting some ballpark figures. Eg http://www.jaycar.com.au/productView.asp?ID=HH8546&CATID=&keywords=heatsink&SPECIAL=&form=KEYWORD&ProdCodeOnly=&Keyword1=&Keyword2=&pageNumber=&priceMin=&priceMax=&SUBCATID= has a thermal resistance of 0.552 degrees per watt. Ask it to dissipate 7W and it will rise 3.5 degrees. Probably overkill for your application but it shows the general idea.

This one http://www.jaycar.com.au/productView.asp?ID=HH8522&CATID=&keywords=heatsink&SPECIAL=&form=KEYWORD&ProdCodeOnly=&Keyword1=&Keyword2=&pageNumber=&priceMin=&priceMax=&SUBCATID= may not be enough. 7W at 11C per watt is 77C. I'd maybe keep the temp below 50C, and assume a very hot day of 40C only allows a 10C rise.

Look for a heatsink with a thermal resistance of about 1C per watt.

My favourite temp sensor is the LM35. Cheap and works well with ReadADC10. Small enough to strap to the battery pack with some duct tape.

 

[Dippy]

If it's just for battery charging then this is a prime candidate for Switched mode operation.

I realise the circuit is more complicated, but it would still be smaller/cheaper than a chuffing great heatsink. And a 1oC / W heatsink is BIG. And even a 5oC/W is quite big meaning your tight enclosure would have to be a big tight enclosure with big vent holes in it - hence the rise of the Switcher.

And, of course, a much lower 12V supply current.

If the example circuit too heavy-going for you then a switcher down to a suitable voltage would take the strain off your PNP.

If you want power in a small box then I think you only have one choice - apart from buying a charger.

Alternatively a DIY switcher as BB suggested.

 

[BrendanP]

Thanks Doc, I was just flicking through jaycar catalogue looking at thermistors but will use your suggestion instead.

I came across this article:
http://www.batteryuniversity.com/partone-11.htm.

Ive got plenty of HRF3205's here from another project. Massive overkill I think but there will be plenty of margin if I want to use the design to charge larger battery packs.

Yep, Ill go the switcher Dippy. I want this thing to be opperable in all temps (and to have the smarts not to charge below 10C). Using another picaxe will allow more functionality as well. Indicator LED's fast flash fast charge, slow flash trickle. etc.

 

[BrendanP]

Doc have you got the debug figures for the LM35? ie. at this temp the device gives this readadc figure?

Im reading the data sheet now.

 

[moxhamj]

It is 10mV per C. 24C will come out at 240mV but there are several things that alter the final numbers, including the regulation of the 5V supply which has to be quite precise. Even then I'm still not sure what is accurate as I sit in front of two digital and one mercury thermometer all of which seem to disagree by a few degrees. Even my two multimeters disagree on the mV from the LM35. So the best I can get is to say +/-2C. I'm sure a lab has the proper equipment to measure it more accurately.

I did measure these adc10 values for my car radiator fan controller:
' 30 degrees = 61, 40 degrees = 81, 50 degrees = 102, 60 degrees = 122

Just a thought re heatsinks - a diecast aluminium box may work out similar price to a heatsink, and the whole box effectively is a heatsink.

 

[BrendanP]

A metal housing is as you say ideal as a heatsink, problem is I have to get this whole thing into a 40MM diameter tube, Ive already had closures cnc'd and had the main pcb made, the battery charger board piggy backs onto the main board via 2mm bolts. I had to do it like that to keep the lenght down.

 

[moxhamj]

Hmm - 7W into a 40mm plastic tube is going to be hot whichever sort of heatsink you use. Plastic tends to get soft when it gets hot. Is it sealed? Can you add a small fan like off a CPU?

 

[BeanieBots]

7W in 40mm plastic.
Throw out your PCB and start again with a buck switching regulator.
Experiment first, even a good switcher will get quite warm unless you take care. Schottky diodes even give off around .5Watt/Ampere.
Wind your own inductor. You will be able to reduce the I^2.R losses (heat) by using thicker wire than commercially available inductors.
In such a small (and presumably draft free) volume, everything needs to be taken into account, not just the heat generated in the main transistor. Don't forget to de-rate the caps. Not only will they get hot from the current passing through them, but they will also be heated by other local power losses.
Given your constraints, you need to concentrate on not producing heat rather than working out how to get rid of it.
Have a look at commercial heatsinks. Even 40mm copper tube will get hot trying to conduct away that amount of power!

 

[BrendanP]

Thanks BB, I came to similar conclusions aftrer further thought. Once approach is to do the V reg stuff external to the tube which is how I'll probabaly have to go.

 

[BeanieBots]

Maybe for this particular project, but have a go with SMPS.
It could save you many problems with future projects, and is a great way to learn about noise and what to do about it.

 

[BrendanP]

I will do so BB, I can use what I know now and once I get familiar with DIY SMPS I can use that as an add on to the main board without having to redraw it. Becasue of space constraints the MAX IC sits on its own baord which piggie backs onto the main board via 2mm bolts which has the picaxe ,GSM module etc. on it.

Just got email back from laser cutter to make disks to hold board, batts, antenna etc. in tube. My brain hurts............

I'll post or email you BB re. the SMPS as I have queries if thats OK.

 

[BrendanP]

Attention BB

You'll recollect I posted a while back re. some of the problems I was having with heat dissipation using a linear battery charger bassed on the MAX712.

I've made a proto of the device using the linear charger, I can get away with the linear in this case because Im charging at a very slow rate to keep the heat down.

The linear is only good for the proto. I need to get switch mode sorted. The MAX713 has functionality that I don't need such as temp bassed termination of charging. Room on the pcb is very tight so I wont to keep the BOM count down as much as possible.


At the time of my original post BB suggested building my own charger using a FET, a inductor, diodes and some caps. I can see a set up along these lines in part of the switch mode circuit depicted on the mAX713 data sheet page 16 which is here.

http://datasheets.maxim-ic.com/en/ds/MAX712-MAX713.pdf

I dont undertand fully how such a circuit would work. I presume the idea is that a picaxe turns the FET off and on via PWM and this regulates the current . The caps smooth out the noise, at least one of the diodes, that in between the inductor and then battery + is there to control the direction of current flow.

Ive googled to try and find some suggested DIY circuits but no joy. Any advice input much appreciated.

 

[Ralpht]

Dr_Acula - It is 10mV per C. 24C will come out at 240mV but there are several things that alter the final numbers, including the regulation of the 5V supply which has to be quite precise. Even then I'm still not sure what is accurate as I sit in front of two digital and one mercury thermometer all of which seem to disagree by a few degrees. Even my two multimeters disagree on the mV from the LM35. So the best I can get is to say +/-2C. I'm sure a lab has the proper equipment to measure it more accurately.

I did measure these adc10 values for my car radiator fan controller:
' 30 degrees = 61, 40 degrees = 81, 50 degrees = 102, 60 degrees = 122

Just a thought re heatsinks - a diecast aluminium box may work out similar price to a heatsink, and the whole box effectively is a heatsink.



The above is pretty correct and the box as a heatsink is a very good idea - saves space and cost.

Dr_Acula - there is an old saying :

A man with one watch always knows what time it is - a man with two is never sure ....

Unless your multimeters (and thermometer in your case) are fully calibrated you will always have measurement discrepancies. Use only one of them and trust it. I've always found that Fluke meters are pretty spot on - just cost a lot.

 

[moxhamj]

re "I dont undertand fully how such a circuit would work." There are two sorts of switching regs - buck and boost. Boost increases the volts and decreases the amps. Buck decreases the volts and increases the amps. This one is a buck.

Referring to fig 19, I'm not 100% but I think D1 is just there to prevent the batteries discharging through the fet if the batteries are present and the circuit is not powered up.

D2, L1 and M1 do all the clever work. I like to use a water analogy for inductors. Flow is amps, and pressure is volts. If you have a long big pipe and you turn it on then it takes a while for the flow to start up. Perhaps the best example is when someone drives over a fire hydrant in a movie - the water geyser doesn't instantly appear. Similarly if you suddenly turn off a tap the water wants to keep flowing, and you get water hammer in the pipes.

An inductor is like a long pipe, or maybe even think of it as a pipe curled up on a hose reel. You turn on the flow (turn on the fet) and the flow slowly starts (current slowly starts) and it slowly builds up. So the current is going through the fet, round and round the inductor and into the load. The current is rising slowly because the inductor is storing energy. Then the fet is turned off. Now the circuit is drawing no current from the mains, but the energy in the inductor is still going round and round and it has to go somewhere. It keeps going in the same direction, and it needs to draw some energy from somewhere, so it "sucks" some energy up from the ground, through D2, through the inductor and through the batteries. It is as if the inductor has become a bigger battery which is now charging the rechargeable batteries.

So this isn't just switching power on and off. It is converting it from one voltage to another. You might start off with 20V at 100mA and end up with 5V at 400mA. (minus the losses which might be 10%).

Google "buck" and "boost" circuits. There are some other things you need to know about, in particular the fact that inductors can saturate, which means they cannot carry any more current. Building your own is tricky because you want to run close to saturation but not actually go into saturation.

RalphTeichel - I like that quote. Of course, a man with one watch running slow will never know the correct time, but a man with a stopped watch will at least have the correct time twice a day. I'm off to take the batteries out of my multimeters Mine are cheapies and so very likely are out of spec. Also the battery in one went flat two days ago, so those readings were likely off due to a nearly flat battery. Might look at a fluke. Have always wanted one.

 

[Ralpht]

I also have a Jaycar cheapie and my Fluke. The cheapie is used when I'm away from the bench and likely to be rough a bit.

Most of the mid to higher range Flukes are calibrated and you can be sure that they will be very accurate and will last forever. Depending on the model you choose, don't choke too much when you see the pricetag.

Whether you really need a meter that accurate, for hobby use, is purely a personal thing. The cheapies will do the job quite nicely.

Sort of like the difference between owning a Ford or a Merc. Both will do the job, one just feels better .......

 

[BCJKiwi]

Had issues with a failing car battery a few weeks back and had it replaced. The new one did not seem to be charging properly so went back the Auto Electricians who had replaced the battery.

The Tech got out his special Auto Electrician's Model Fluke Multimeter and was very disparaging about the voltages I had recorded with my 'Cheapo' DMM from Jaycar (a QM1538 datalogging model) - pleased to note that the voltages were within 0.01V of the Fluke!

It seems the regulator in the Alternator is set a bit low as it seems to maintain charge even on short running but at a good 0.5V less than expected. Also took a few decent charge cycles with a new-fangled electronic battery charger to get the Battery stabilised.

 

[BrendanP]

I got a Meterman DVM from Dick Smith a while ago for under a $100. Im very happy with it.

 

[Dippy]

Yes, in a home-made buck converter the diode just stops reverse current as said by Dr_Ac.

A home-made one is quite easy, but the 'proper' ones will operate faster. I don't mean necessarily the PWM speed, but the feedback/correction speed.
You may find home-made ones a little noisier but I'm sure with a little care you can live with it.
Max vs Home-Made?
Well the MAX is designed for the job but I must say looks heavy on the BOM. I'm sure there must be simpler ones around.
Obv. you'll have to spend time writing code and one more thing; I realise that this was covered weeks ago but just to reiterate - you WILL need a MOSFET driver for fast PWM if you want cool running i.e. High Efficiency and proper SMPS type switching. The types of driver and reasons for using them were covered in detail some weeks ago so I won't repeat it.

Multimeters:
Well, your Keithly and Flukes are the Rolls-Royce.
TTi and Metrix are the top of the range Mercs.
Meterman , B&K Precision and numerous non-Chinese others are your Ford Mondeos. i.e. some look superb value and if you get a good one it could give a Fluke a run for it's money.
Anything Made in PRC are your 70s Skodas. i.e. they may work fine and for kids they're good enough, but would you trust it and would it last? You may be lucky...

Question: If you were to buy just one DMM, which would you trust straight out of the box; A Fluke or a £5 Ebay anonymous special offer. If the latter then you obv don't take electronics seriously.

 

[BrendanP]

Thanks Dippy, I've managed to jam everything on the board as per the MAX 713 data sheet. I've gone half SMD and half thru hole so its helped out with the space problem a bit. Ill proabably just go with the MAX design for now and experiment at my leisure

Im kinda under the pump with this thing, I've spent $16,000 on advertising for this device which comes out 8th August, so I better have it all going OK with product to sell by then.

Yeah I know its madness to rack up so much debt bassed on a proto. Everything else is under control, well sort of, there are some flow controll issues with the GSM module.......

Im with you on the tool issue Dippy, I had my Husqvarna chainsaw out today, quality tools are a joy to use. I've got a Hitachi router I bought over 20 years ago that is still made today. Mine still runs beautifully. I've never regreted spending money on rifles or tools.

I want to get a desoldering station, any recommendations?

 

[Dippy]

And ammo too I hope. I bought some cheap 22 years ago and now and then they jammed right in the ejector. Changed to Eley and no probs. But those days have gone in UK.

Desoldering station? Dunno, the last proper one I used was ages ago and it was a Weller (now Cooper Tools).

 

[BrendanP]

Ahh the glory days in the early 80's when you could get the local rag in Melbourne and sally forth and buy a SLR or AR15 from some kindly soul out of the classifieds..... It used to be like 'Appocalypse Now' when we'd go out on shooting expeditions with bulk lots of ex mil FMJ's, (no stoppages with that stuff).

I've got a weller/cooper soldering station that has given good service I'll have a look at thier desoldering gear.

 

[eclectic]

@Dippy.
As an aside, I've just looked at the UK Farnell website.
Clearance section.
Fluke 1587 + bits = £95 +VAT.

Somehow, my typing fingers were taken over.
I'm sure I won't regret it.

Brendan. Sorry for the interruption. Back to the plot.

e.

 

[Ralpht]

eclectic - As an aside, I've just looked at the UK Farnell website.
Clearance section.
Fluke 1587 + bits = £95 +VAT.


Not too bad a model. Still here in Oz it wil translate to something approaching $300.

I obtained mine from the "blue supermarket ", - Royal Australian Airforce - they were upgrading a lot of test equipment and normally old stock goes to government auctions. I managed to wrangle a deal and got it for a song.

I was only half way through the song when they stopped me and said they considered themselves paid more than enough ......

The joys of being in the military and having access to stuff you normally can't in civilian life.

 

[Dippy]

Haha. Yes, a friend of mine (alledgedly) got 3 brand new variacs which the MOD were upgrading too