Maker Pro
Maker Pro

SOT23 PNP is dissipating too much?

Status
Not open for further replies.

eem2am

Aug 3, 2009
429
Joined
Aug 3, 2009
Messages
429
Hello,
Is the following PNP transistor going to overheat in this circuit (attached)?

The PNP transistor is put there so as to provide a low output on its collector if the load ever goes open.
However, the PNP is just a CPH3105 type, and is in a SOT23 package, and is on a minimal footprint on the PCB, with no cooling copper added to any of its pads.
This PNP is dissipating 175mW (Ieb * Veb)

CPH3105 PNP transistor datasheet:
http://www.onsemi.com/pub_link/Collateral/EN6084-D.PDF
 

Attachments

  • PNP circuit.jpg
    PNP circuit.jpg
    35.9 KB · Views: 149
Last edited:

Arouse1973

Adam
Dec 18, 2013
5,178
Joined
Dec 18, 2013
Messages
5,178
Mine dissipates 15mW, cant see the problem. :)

Adam
 

Attachments

  • PNP.PNG
    PNP.PNG
    4 KB · Views: 157

eem2am

Aug 3, 2009
429
Joined
Aug 3, 2009
Messages
429
175mA * 0.1K = 17.5V.....that's the problem with your circuit...
I know , you are joking....sorry, temporary sense of humour failure on my part.
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Jan 21, 2010
25,510
Joined
Jan 21, 2010
Messages
25,510
I know , you are joking....sorry, temporary sense of humour failure on my part.

No, I think he fixed your circuit. (It's your series resistor value, not his)
 

eem2am

Aug 3, 2009
429
Joined
Aug 3, 2009
Messages
429
so you say just reduce the series resistor to say 7R5?
I see your point, but the manager doesn't want to change the circuit as it is....and the 100R is an 0603 footprint...so I couldn't just reduce the 100R to 7R5, without changing the resistor footprint to a higher power one.......and they don't want any changes.
If I can assure them that the PNP will overheat in the original circuit, then they'll make the change......I always take sot23 on minimal pads as 263degC/Watt...do you agree?.........
I also take Vbe (max) for CPH3105 as 1.2V, therefore, temperarture rise in this sot23 PNP on minimal pads is 1.2 x 0.175 * 263 = 68degC, and with an internal ambient of 60degC, this gives a T(junction) of 126degC, which is too hot...do you agree?
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Jan 21, 2010
25,510
Joined
Jan 21, 2010
Messages
25,510
I'm surprised that the transistor allows such a large base current.

With that in mind, reducing the value of R2 to 9,2 ohms would probably give you about half the current through the transistor and the resistor.

It might mean that the transistor does not overheat.

If you place an 0805 resistor on the 0603 pads (you should be able to do it) then you might have a component able to withstand the dissipation.

It's an awful design though. I just hope you're not responsible for it. Whoever thought of carrying the load current through the base-emitter junction should perhaps have come up with something better.

I wouldn't be happy about the thermal stability. The BE junction will hog the current as the temperature goes up. It won't be pretty.
 

Arouse1973

Adam
Dec 18, 2013
5,178
Joined
Dec 18, 2013
Messages
5,178
175mA * 0.1K = 17.5V.....that's the problem with your circuit...
I know , you are joking....sorry, temporary sense of humour failure on my part.

If your going to joke then at least be funny, and oh draw a circuit that works.
 

Arouse1973

Adam
Dec 18, 2013
5,178
Joined
Dec 18, 2013
Messages
5,178
175mA * 0.1K = 17.5V.....that's the problem with your circuit...
I know , you are joking....sorry, temporary sense of humour failure on my part.

You obviously don't understand the circuit, it is quite complex. The circuit is drawing 175mA, 165.8mA through the base junction and 9.2mA through the resistor. So it is dissipating 175mW as you mentioned. I thought I would fix this for you as you were obviously struggling.

If your trying to be clever, next time get it right.

Adam
 

eem2am

Aug 3, 2009
429
Joined
Aug 3, 2009
Messages
429
Yes I knew you were suggesting reducing the resistor value and shovelling the current through that instead of the BE junction.
I honestly thought you were kidding as you left the 100R in there, and added the smiley face, and indeed your circuit, did work, as such.
I actually thought you were (understandably) taking the Mick out of the circuit's "pathetic-ness", as it certainly is pathetic, and the guy who designed it has legged it and left us with it.......our boss doesn't want to change it unless we can absolutely prove to him that the circuit is a "dog".
I seek that proof.

If I told you which car this circuit is being used in..you simply would not believe me in a million years, -you really would take the Mick...you simply wouldn't believe it but its true.

Steve's modification does the job of passing the 175mA, however, the 9R2 resistor will dissipate somewhere up to 280mW..... but the transistor will dissipate more, the more the heat rises, as its vbe gets less.

Id have to change to a 1812 resistor size to use the 9R2 resistor.....and the boss doesn't want changes...unless I can "prove" to him that the circuit of the top post has a "death curse" hanging over it.

My "death call" for the circuit of the top post , centres around the following reasoning that a SOT23 package cannot be allowed to ever dissipate more than an 0805 resistor (125mW)....

SOT23 PACKAGE DISSIPATION SCIENCE:
Regarding the allowable dissipation of an sot23 device...
Comparing it with SMD resistors is helpful..
A sot23 (3mmx1.3mm) body size is slightly smaller than a 1206 resistor (3.2mmx1.6mm).
A SOT23's pad contact area is, however, a mere 1.2mmx0.4mm.
A 1206's pad contact area is 1.6mm x1mm....much more than a sot23.

...so we can say that a sot23 wouldn't be expected to have a power rating as much as a 1206 resistor (250mW)

In fact, the pad area for a 0805 resistor (1.25mm x 0.8mm) is more than for a sot23.
An 0805 resistor, does however have a smaller body area(2mm x 1.25mm) than a sot23.

So I would say that a sot23 could dissipate no more than an 0805 resistor, i.e. 125mW.

At my last company, a Consultant Engineer, who runs his own electronics consultancy, has a range of lighting electronics products on sale to the commercial lighting world, and has completed a number of electronics projects for the military, plus used to design the neon lights for picadilly circus, plus used to design audio amplifiers for commercial sale, told me never to dissipate more than 100mW in a SOT23 package. (He was referring to my use of SOT23 BJTs in a battery charge circuit., for one of his sister companies)

Is there concurrence with the 100mW maximum for a sot23?

Is there at least concurrence that a sot 23 is at most, equivalent to a 0805 resistor in terms of dissipation allowance.?
Its worth noting that resistors don't ever come in SOT23 packages....and this suggests that the SOT23 package is not as good at dissipating heat as a SMD chip resistor package.?
 
Last edited:

Arouse1973

Adam
Dec 18, 2013
5,178
Joined
Dec 18, 2013
Messages
5,178
Why don't you just read the data sheet for the device. They said once computers would not go higher than 680K. Got that a bit wrong did'nt they. Not with DOS anymore, do you remember DOS?
 

Arouse1973

Adam
Dec 18, 2013
5,178
Joined
Dec 18, 2013
Messages
5,178
Let's hope its not. A BMW or Honda. But I doubt that very much
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Jan 21, 2010
25,510
Joined
Jan 21, 2010
Messages
25,510
Steve's modification does the job of passing the 175mA, however, the 9R2 resistor will dissipate somewhere up to 280mW

Actually the idea was to split the voltage between the transistor and the resistor, with each dissipating around 0.07W (this is due to a guestimated Vbe of 0.8V at 88mA

but the transistor will dissipate more, the more the heat rises, as its vbe gets less.
The question arises in my mind... will it?

I calculate the dissipation o the transistor in that configuration, carrying 100mA to be 0.069W (as opposed to 0.070W at 87.5mA)

I think you should examine the thermal issues and see if my original expressed skepticism about the thermal stability was in fact correct. Fresh eyes on this suggest to me that I may have been wrongly concerned.

A graph showing the power dissipation of the resistor and transistor at varying Vbe, and assuming a constant 175mA might be instructive.

Id have to change to a 1812 resistor size to use the 9R2 resistor.....and the boss doesn't want changes...unless I can "prove" to him that the circuit of the top post has a "death curse" hanging over it.
I doubt you'd have to use a resistor that large, however it would be wise to consider the thermal environment (hint: derating) as well as the duty cycle.

Perhaps (dare I say it) some practical tests might be instructive -- especially since the datasheet doesn't give Vbe over voltage and temperature.

SOT23 PACKAGE DISSIPATION SCIENCE
That is conjecture (although if you had no data, it could be a starting point). Read the datasheet to determine the actual story. The board material, temperature, and amount of copper connected to the device will all have a bearing.

Remember that there are resistors that can dissipate more power than others. It may be cheaper to specify a "better" part than to redesign the board.

Is there concurrence with the 100mW maximum for a sot23?
Perhaps that's a good rule of thumb, but that's all. You will note that my design falls within that (which is more due to coincidence than by effort on my part)

Is there at least concurrence that a sot 23 is at most, equivalent to a 0805 resistor in terms of dissipation allowance.?
Is there concurrence that you should read the actual datasheet and not make stuff up for yourself?

Can you point us at the datasheet for the type of resistor currently in use? (If you can't, you should be ashamed of yourself)

Its worth noting that resistors don't ever come in SOT23 packages....and this suggests that the SOT23 package is not as good at dissipating heat as a SMD chip resistor package.?
That's no more worth noting than that LEDs don't come on TO-220 packages, and then concluding that TO-220 packages are not as good at passing light.

And in any case, resistors DO come in sot23 packages.
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Jan 21, 2010
25,510
Joined
Jan 21, 2010
Messages
25,510
I must have had DOS on my ZX81, but I don't know.

Am I supposed to extract meaning from this?

The datasheet for SOT23 says that at 70degC ambient it can handle 230mW on standard footprint..
http://www.ricoh.com/LSI/product_power/pkg/sot-23-3.pdf

Yeah, but that's a little pointless isn't it? I mean, you have the datasheet for the device in question. Why not look at that?

In any case, you are operating that transistor in a mode that is not anticipated by the datasheet. I have no idea what exactly will happen. And I guess that you don't either.

That's why I suggest you extract your engineering brain from the comfy chair in front of the computer and take it somewhere where you can do an actual test.

I'm not going to do one for you this time. I'm getting pretty sick of your pontificating based on inappropriate datasheets.

...but it doesn't seem right as an 0805 resistor, which has twice the pad contact area, is only rated to 125mW at 70degC.

*THINK*

Have you looked? Here for example.
 

eem2am

Aug 3, 2009
429
Joined
Aug 3, 2009
Messages
429
Can you point us at the datasheet for the type of resistor currently in use?
thanks for advice... BTW the designer has legged it, and the in-house engineer is too busy as yet to give me the bill of materials, and it doesn't say on the schematic, other than its an 0603 resistor....the electronics lab is at another site, where I am not allowed to go.....I have just been employed for a few weeks to check over the pcbs & schematics before they go into full production.
I suggested doing a test in the lab but they fobbed me off, saying that they powered it up and it seemed to work ok, so what was I worried about....
The CPH3105 datasheet , as you allude, is very non-informative in this situation.
I'll have a look at your Sfernice link though.
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Jan 21, 2010
25,510
Joined
Jan 21, 2010
Messages
25,510
Are you inventing problems for people again, or does this circuit actually fail in production?

What lights are we talking about?
 

eem2am

Aug 3, 2009
429
Joined
Aug 3, 2009
Messages
429
here is the overall circuit....I put the led driver and leds as just a resistor as I cant put all that lot in the simulator.
The leds are driven by LDU0830S350 modules.

its reverse/fog lights at the back of a car......a fog or reverse is 2 series leds with 350ma in them.......so what you see there is a right side fog and a right side reverse driver......the fog goes through one pnp, and the reverse through the other.

the circuit does not fail yet, but its fresh off the line, and has only just last week been fitted into the first car.......they probably haven't even turned the fogs on yet.......thee are seven more cars coming.......then more, and then they plan to re-use the circuit for other cars.......the circuit hasn't been tested in a thermal chamber yet...just run on the lab bench
 

Attachments

  • switch damage.pdf
    14.2 KB · Views: 69
Last edited:

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Jan 21, 2010
25,510
Joined
Jan 21, 2010
Messages
25,510
eem2am, I am locking this thread now.

If you want it reopened, you need to PM me with answers to ALL of the following questions:

1) Is it one of your duties to check this design?

2) Have you been asked to check this design?

3) Are you responsible for the correct operation of this design?

4) Have there been any failures from this unit?

5) Have there been any indications of potential failure of this unit?

6) Do you have access (and have read) all of the design and testing material for this unit?

7) Do you know for certain that "they [...] haven't even turned the fogs on yet"?

8) Do you remember anywhere in your past any occasions where you have raised red flags about poor design? How did that work out for you?

Questions answered in a PM, numbered for my convenience please.
 

KrisBlueNZ

Sadly passed away in 2015
Nov 28, 2011
8,393
Joined
Nov 28, 2011
Messages
8,393
Please excuse me Steve while I post on this closed thread.

A few comments on the design. These are in addition to Steve's comments in post #12 which you should also read carefully.

The CPH3105 is specifically rated for a maximum base current of 600 mA, and this suggests to me that it's designed to monitor load currents directly, without a separate path for most of the load current. But...

C1 and C2, 22 µF capacitors from Q1 and Q2 bases to ground, and the smoothing capacitors in the switching LED drivers, will cause a large current spike in the base-emitter junction of Q1 and Q2 when power is applied. Depending on the rate of rise of the voltage on M1's drain, this spike could easily be enough to damage Q1 and Q2.

Characterising LOAD1 and LOAD2 as "175 mA" is a misleading oversimplification. They are actually switching converters, and their current draw will vary over a moderate range in roughly inverse proportion to their supply voltage, so they have a negative resistance characteristic, so they're neither a resistance nor a current sink. I understand that you can't easily model them in the simulator, but any or all of those factors could be important to us humans who are trying to understand the setup, and should at least be shown as comments on the schematic. Please bear this in mind for future times when you are about to redact potentially important information from your schematic and/or your description. We should not have had to wait until post 18 before discovering this important information.

As for the issue of continuous power dissipation in the transistors, I think it's worth investigating. You didn't mention this, but Q1 dissipates extra power due to the emitter-collector current that feeds LOAD2. This is a lot less than its base-emitter dissipation, because Vce(sat) is typically only 100 mV, but it still exists.

The CPH3105 data sheet gives a maximum Vbe of 1.2V at a base-emitter current of 100 mA and a collector current of 1A. Q1, and especially Q2, run at a much lower collector current, and collector current does affect the Ib vs. Vbe relationship, but not to a huge extent, and as far as I can tell using LTSpice with common transistor types, reducing Ic will also reduce Vbe for a given Ib, so you would be erring on the safe side. Assuming LTSpice's transistor models are accurate in this respect.

As for the actual maximum continuous load currents, they will occur with (a) minimum supply voltage, (b) maximum voltage drop across L1, D4 and M1, (c) maximum base-emitter drop in Q1 or Q2, (d) maximum Vce drop in Q1 (for LOAD2), (e) minimum efficiency in the LED driver (95% in the table in the data sheet, but this is not guaranteed), maximum LED current (6% high for an LDU08xxxxxx), (f) maximum LED forward voltage at that current, and (g) worst case temperature extremes for all of these. I don't know most of those figures; you should work them out. In the meantime I would design using load currents of 250 mA AT LEAST, preferably 300 mA.

Ignoring the small current through R2, and assuming that the 1.2V maximum Vbe at 100 mA also applies at 300 mA (NOT a valid assumption), the worst case Q1 dissipation would be (0.3 * 1.2) + (0.3 * 0.1) = 0.36 + 0.03 = 0.39W, say 400 mW. Based on any of the thermal resistance figures I've seen in this thread so far, that is a problem.

These calculations all relate to steady state conditions. The possibility of damage due to capacitor charging currents at switch-on is a separate matter and could still be an issue even if the steady state dissipation is within limits.

I would use a different approach. First, if possible I would avoid monitoring the load current in the first place! That's just because I don't know why it's necessary. If that's unavoidable, I would insert a proper diode (e.g. 1N4001) in each load path, and add a tiny amount of resistance somewhere (e.g. in series with D4, which is only rated for 1A continuous anyway), to minimise the turn-on current. Then monitor the drop across the diode with a transistor operating at low current. (You may need some small components to ensure enough bias voltage under all conditions.)

Other general comments:

That seems like an awful lot of circuitry. Apart from the overvoltage protection (D2, D3, Q3) and the pass MOSFET, what functionality is actually needed? Why is it necessary to turn the pass MOSFET OFF if either of the loads goes open circuit?

What is the purpose of D7, Q5 etc? Are these just to ensure that Q4 and the pass MOSFET remain ON for a short time at power-up? They will also have an effect if the input voltage is low; is this not a problem, or is it intended behaviour? If the latter, what is the reasoning behind it?

Is this an exact copy of the circuit in production? If you're about to answer YES, please think again. I doubt that a designer would use two zeners in series (D2, D3) when one is available in the right voltage; I assume this is a change you made because of LTSpice's limited component options. I have done exactly the same thing myself. What other changes have you made for this and other reasons? Is M1 really an Si7113DN in the production circuit? If not, please add an indication of the actual part number, the way you did for Q1 and Q2. Are there any other differences? Please look carefully.

In this application, are the fog lights and the reversing lights actually always driven at the same time? They have different purposes, don't they? Is there another part of the circuit that you haven't shown? If they're always driven together, why not use a single LED driver and connect all four LEDs in series? Is it because the fog lights and the reversing lights are too far apart? Or because there is not enough voltage available from the drivers for four LEDs in series?

eem2am, you can post any response to this once Steve reopens the thread, or PM them to me if you prefer.
 
Last edited:
Status
Not open for further replies.
Top