LM7812, LM7912, diode bridge operating temperature

[ballen]

Dear EEVbloggers,

To start off: I'm new here.  I hope this question is in the right forum, if not, please tell me where to move it.

I've got a 1990s era precision clinometer (0.1 arcsecond resolution, 0.2 arcsecond precision) which has both AC line power and built-in NiCd rechargables.  I'm replacing each of the two NiCd batteries (9v, 1100mAh) with a pair of 18650 Li-ion batteries and battery management/balance  controller board.

The original charging circuit for the 9V NiCds was lousy: a 39 ohm 2W resistor in series with the +-12V regulated supply and a 0.7V protection diode drop.  The intent was that when the NiCd voltage was ~7V, this would provide about 100mA of charge current, reducing to a 25mA trickle when the battery had gotten up to 10V.

I've replaced the 39 ohm resistor with a constant-current source made from an LM1117-adj (low drop-out LM317) plus reverse protection diode.  I've tested that, and it works well.  Right now I have the current source set to provide around 400mA, which charges the Li-ion batteries in ~5 hours.

My only concern is the following.  The designers intended their +-12V supply to provide ~150mA to power the meter plus the charger.  Right now, I am drawing close to 450mA when the batteries are charging.  I want to make sure that I don't do any long-term damage.

The +-12V supply is simple: a center tapped ring transformer, full-wave rectifier, then a 7812 and a 7912 regulator, TO-220 package, each on a finned heat-sink, about 20mm square and 20mm high.  The transformer is rated for 30 or 40VA, so no issue there.  The supply circuit is fused before the regulators and after, 1A slow-blow, again suggesting that 450mA is not too much.

As far as I can tell, the only critical components that I need to worry about are the bridge rectifier and the 7812/7912.  DC coming in is about +-15V, so the regulators need to dissipate about 1.4W each. I put a thermocouple on the 7812 (which is providing slightly more current than the 7912) and the tab's steady-state temp in charge mode is about 70C, which would put the chip around 80C.  The bridge is slightly too hot to keep my finger on it, so probably running around 75C. (It's one of those little round packages and there is air space around it, so I can probably stick on a round finned heatsink and drop that a bit.)

My question for the forum is this: am I OK running the bridge and 7812/7912 at these temps?  Or will it cut their lifetimes enough that I should either drop my charging current (say to 250mA) or try to fit larger heat sinks to the regulators?  I know that these are rated up to around 125C junction temperature, but I'm sure it's bad practice to come any where near to that.  How far away should I stay?

By the way, the actual metering part of the instrument runs on +-5V internally.  The +-12V supply and the +-9V (nominal) battery are only there to power LDO +-5V regulators downstream.  Also, the sensitive/precision/low-offset/balanced/low-drift stuff is on the other end of the board, well away from the heat of the power supply circuitry.  So the instrument function is remarkably unaffected by these mods.  It's high-end gear (Taylor-Hobson Talyvel 4) and it's my impression that the designers left a lot of headroom.

Cheers,
    Bruce

[SilverSolder]

Personally, I'd just go with the 250mA charge current idea unless there is some reason you need more?

[ballen]

Personally, I'd just go with the 250mA charge current idea unless there is some reason you need more?

I'd like to be able to charge the batteries faster if possible.  So really would like to hear about what operating temperatures are reasonable for the LM7812, LM7912, and diode bridge.

[Paul Moir]

The old adage is "if it's too hot to touch, it's too hot", which admittedly is an overly conservative and not real engineering.  The fact is the regulator's thermal shutdown won't kick in until 150degC and if your tab is 70, then the die is probably 80-90 meaning you still have a tonne of headroom.  Diodes are capable of running very hot with board damage being a real limitation.
But being overly conservative I would probably try to find some better performing heatsinks and put something on the bridge rectifier if it was easy enough.  Just going from bright finished to black anodized is probably enough.
EDIT:  Note the datasheet has a recommended top die rating of Tj=125degc.

[mikerj]

I've replaced the 39 ohm resistor with a constant-current source made from an LM1117-adj (low drop-out LM317) plus reverse protection diode.  I've tested that, and it works well.  Right now I have the current source set to provide around 400mA, which charges the Li-ion batteries in ~5 hours.

With no other voltage limiting, when the charging terminates you'll have around 11v on the input to the BMS, is that within it's specs?

The 7812 temperature is not disastrous in terms of reliability of the regulator itself, but consider whether ambient temperature is ever likely to rise e.g. using the device in the middle of summer and also if there are electrolytic caps near the heatsinks that are likely to suffer premature death from high temperatures.  With so much headroom to the 5v regulators you may be able to add some series resistance to drop the 7812 input voltage when charging.

[ballen]

I've replaced the 39 ohm resistor with a constant-current source made from an LM1117-adj (low drop-out LM317) plus reverse protection diode.  I've tested that, and it works well.  Right now I have the current source set to provide around 400mA, which charges the Li-ion batteries in ~5 hours.

With no other voltage limiting, when the charging terminates you'll have around 11v on the input to the BMS, is that within it's specs?

Yes, it's fine. The BMS has a pair of MOSFETS which shut off any current flow to the unit (and the batteries) once the Li-ion cells reach 4.2V each (total 8.4V).  It's perfectly happy to see 11.2V at the input when no current is flowing.

The 7812 temperature is not disastrous in terms of reliability of the regulator itself,

I'm happy to hear that.  Is the MBTF at ~90C junction temperature still thousands or tens of thousands of hours?  (It follows from the silicon junction bandgap  voltage that each 10C increase in temperature will double the failure rate or decrease MTBF by a factor of two.   So if I had an MTBF number at some nearby temperature, I could easily guestimate the MTBF at my measured operating temp.)

but consider whether ambient temperature is ever likely to rise e.g. using the device in the middle of summer and also if there are electrolytic caps near the heatsinks that are likely to suffer premature death from high temperatures. 

Worst-case ambient could get up to the high 20s, but not 30s or 40s.  So that's OK.

The PCB is not crowded around the 3-pin regulators.  The closest electrolytic is 2cm away, and I've replaced it with a low-ESR 105C-rated capacitor.  (I needed to get a much higher voltage rating to match the package size and lead spacing from 30 years ago!)

With so much headroom to the 5v regulators you may be able to add some series resistance to drop the 7812 input voltage when charging.

It's a good idea, but I don't think I can do that.  My current source is an LM1117-ADJ which has a 1.2V drop out plus another 1.25V from ADJ to OUT.  So tracing the circuit that charges:

-  start at 12V from the 7812, drop 0.7V in a protection diode to 11.3V.
- Then another 2.5V drop in my current source brings me to 8.8V.

That's just enough to run the BMS and charge cells.  I've only got 0.2-0.3V of headroom for charging.

I also thought about taking my charger power from the stage before regulation.  But I don't want to tear up the circuit topology that much.

EDIT: I just found MTBF data on a TI web page.  They estimate about 3 failures per 10^9 hours at 55C junction.  So if I increase this by four x ten degrees, I should get a lifetime shorter by a factor ~ 1/16.  That still gives 20 million hours -- I can probably stop worrying about this.

[trobbins]

Diode dissipation design is typically skipped over, especially for 1A av type diodes, but there is a subtle disconnect between datasheet rating for Iav and when such a diode is used for common garden mains rectification with a capacitor first filter.  If you are keen to 'bridge' the design gap then the linked article has a section on that issue:
https://www.dalmura.com.au/static/Power%20supply%20issues%20for%20tube%20amps.pdf

[ballen]

Diode dissipation design is typically skipped over, especially for 1A av type diodes, but there is a subtle disconnect between datasheet rating for Iav and when such a diode is used for common garden mains rectification with a capacitor first filter.

I read the article that you linked, but don't see the relevance.  It shows that the simplest calculations of power dissipated in a diode can be inaccurate by a factor of up to two, because there is a larger than estimated/expected voltage drop across the junction.  But I don't need to calculate the power dissipated to predict the temperature.  Instead, I can measure the temperature (around 70C).  That does not seem large enough to reduce the lifetime below millions of hours (since a single diode junction is going to be more reliable than an IC at similar temperature).

[trobbins]

As stated in the post, the reference was pointed to if you wanted to extend the design path to come up with an expected junction temperature for your application.

If you only want to use a temp measurement of the external surface of a small package to estimate a junction temp then that's up to you.  Does your bridge diode datasheet have an Rja value (?), as I've only seen Rja identified for larger bridges (which usually have thicker legs), or an Iav value specified for a particular ambient temp for a particular mounting arrangement (like mounted direct to a pcb or some other thermal definition).

[ballen]

As stated in the post, the reference was pointed to if you wanted to extend the design path to come up with an expected junction temperature for your application.

If you only want to use a temp measurement of the external surface of a small package to estimate a junction temp then that's up to you.  Does your bridge diode datasheet have an Rja value (?),

I just looked up the datasheet.  The bridge is rated for 2A continuous without any heat sink, at a package temperature up to 150C.  So pulling 450mA from it (and running at T=70-80C) should be fine.  The forward drop is about 0.7 V so apparently that little plastic package (about 5 x 8mm) can radiate 1.4W of heat without getting too close to 150C.  Remarkable!

[trobbins]

Nice bridge then - do you have a part number, as I'm interested to see what else than W04 style bridges are around?

[ballen]

Nice bridge then - do you have a part number, as I'm interested to see what else than W04 style bridges are around?

W02G, but I got the current wrong, it's rated 1.5A, not 2A.

[trobbins]

It's good to see the Vishay datasheet for W02G range already includes some derating advice for rectifier applications with different Ipk/Iav, and includes a thermal resistance for a particular mounting.  That's much better than the older datasheets.  With a highish Ipk/Iav=20, and some derating for max Tj and say 70C local ambient, each bridge diode can operate with up to about Iav = 0.4A for that default mounting.

[ballen]

It's good to see the Vishay datasheet for W02G range already includes some derating advice for rectifier applications with different Ipk/Iav, and includes a thermal resistance for a particular mounting.  That's much better than the older datasheets.  With a highish Ipk/Iav=20, and some derating for max Tj and say 70C local ambient, each bridge diode can operate with up to about Iav = 0.4A for that default mounting.

I'm puzzled by what you write.  In my application (standard linear power supply) I should have Sqrt(I^2_peak/I^2_average) = Sqrt(2).  Where does "20" come from?

The Vishay datasheet I am looking at has "Maximum average forward rectified current at 0.375" (9.5 mm) lead length at TA = 25 °C" as 1.4A." Where does 0.4A come from?  The "Derating curve" shows a 60Hz resistive or inductive load 1.6A at ambient temp 30C, provided that the bridge is standing 9.5mm off the board on its legs for air circulation and has 5.5mm square pads on the PCB.    Even with a capacitive load and I_peak/I_average of 20, it's still good for 1A at an ambient temp of 25C.

[trobbins]

Rectification in to a capacitor filtered load causes a very peaky current pulse each time a diode conducts.  The configuration of winding impedance, capacitance and load defines how peaky that waveform is, and is easily defined by the ratio of Ipk to Iav.  An Ipk/Iav ratio of 20 indicates a configuration that has a low winding impedance, and a large capacitance - it would be uncommon to have a higher ratio than 20.

The datasheet derating curves are max operating levels where Tj = 150C.  For the 'extreme' configuration where Ipk/Iav=20, the max Iav is 0.98A at 25C ambient.  Derating to an ambient of 70C, gives Tj=150C for 0.6A av, and derating to 125C Tj reduces Iav to about 0.4A.

The local ambient is obvious very dependant on how the equipment is made up.  Sometimes it is easy for the bridge to be located next to a linear reg heatsink, and for that area of a pcb to have restricted air flow, and sometimes little chance of free air flow from an external ambient.  Sometimes the equipment is in a rack, etc etc.

[ballen]

Thanks for explaining.  The bridge diode is slightly too hot to keep my finger on it, so I think there is no significant issue.   I'm convinced that the current draw and power dissipation are not going to reduce its lifetime below millions of hours.  But please tell me if you think that's wrong!

[Siwastaja]

"Hot to touch" is a poor metric. For some people, "hot to touch" means 45 degC, for some others, it means 100 degC, all depending on your personal sensitity and how long you are expecting to keep your finger at the thing. Also the thermal conductivity. Try to hold a +50 degC copper piece in your hands for a minute, can't do it. Touch a 150 degC plastic thing quickly, no problem whatsoever.

Exact answer requires actual engineering, i.e., measure case or heatsink surface temperature, look up Rths of the heatsink, thermal interfaces and finally the device itself (Rthj-c), then calculate actual die temperature, then look up for the maximum allowed on datasheet.

On usual packages like TO220, if you measure the tab is well under 100 degC, all is well; the tab represents the die temperature quite well, and 125 degC tends to be the maximum. Remember to leave margin; if you measure 100 degC in your lab, how about when you run it on a sunny day and the ambient (from the component viewpoint, i.e., inside the case!) is 30 degC hotter?

With the classic round plastic diodes, you may want to measure the leads exiting the diode - they are often thick exactly to remove heat from the diode -, as close as the body as possible, instead of the plastic case which may be quite thick and thermally non-conductive.

Finally, my rule of thumb has been, if a drop of ethanol or isopropanol boils quickly off the device, that's OK,  but if a drop of water evaporates in way less than a second, then it's running too hot.