Linear LED driver works due to emitter degeneration?

[Flyback]

Hello,
This linear LED driver puts 50mA into each string, even though only one string is regulated....and even though the led strings are different lengths.
Is it the emitter degeneration which is equalising the led current in each string?
schem and LTspice sim attached

 

[kubeek]

Try changing the temperature of those transistors.

 

[ronsimpson]

What are the C-B capacitors doing?

The first LEDs are regulated very tightly. 1volt/20 ohms.
The second LEDs should track close. (assuming Q1Vbe approximately equals Q2Vbe)

 

[audioguru]

The transistors are different so why do you think their currents will be the same? They will probably be very different and will be different again if you replace them.

The SIM program does not care that the transistors have different spec's and have different temperatures. The transistor that is driving fewer LEDs has more voltage across it therefore it is hotter and becomes more sensitive and its base-emitter voltage drops increasing the current even more which causes it to heat more which increases the current which causes to heat more which ...

When that circuit is in an IC then the transistors are identical and have exactly the same temperature because they are very close together.

 

[Flyback]

the transistors in the actual application are SOT89's..BCX56-16
**broken link removed**
...As the simulator shows, it doesn't really matter which transistor is chosen, or if they are different, the led currents are pretty much equal.

The transistor that is driving fewer LEDs has more voltage across it therefore it is hotter and becomes more sensitive and its base-emitter voltage drops increasing the current even more which causes it to heat more which increases the current which causes to heat more which ...

...I see what you are getting at, but the one that is regulated is the 2_LED string one, and this is the one that will be heating up more, and its current will try to grow more, but it is regulated , so it will be ok.....the one with 3-leds will thus suffer less current in its leds, which is better than too much current in its leds, so having the regulation in just one string is not ideal, but at least the correct string (the shortest one) has been regulated...would you agree?.......in the actual application, there are ten extra strings of three_LEDs which are like the one shown here, (ie not "directly" regulated).



C-B caps provide a bit of negative feedback and prevent high frequency oscillations

 

[JimB]

U1 in conjunction with Q2 and R1 gives a constatnt current in the collector circuit of Q2.
The value of that current is defined by V3 and R1.
Because of this, the output of U1 is a (fairly) constant voltage.

The output of U1 connects to the base of Q1.
The collector current of Q1 (assuming a reasonably high gain in Q1) is given by:

(Vout_u1 - Vbe_q2)/R2 = Ic_q1

Give or take the variation of Vbe between Q1 and Q2, the collector currents will be the same.

Is this perfect? No.
Is it good enough to drive a bunch of LEDs? - Probably.
Do we need it to be any better? Probably not, but that depends on the application, which we don't know about.

JimB

 

[alec_t]

Is it good enough to drive a bunch of LEDs? - Probably.

I ran the sim with an extra 100mV source inserted in the emitter path of Q1, simulating a shift in Vbe. This gave a ~10% change in Q1 collector current. Probably acceptable?

 

[ronsimpson]

C-B caps provide a bit of negative feedback and prevent high frequency oscillations

For this to work you need to add a base resistor. The time constant is the Rbase and Cc-b. The output impedance of the opamp is too low to really work as a RC time.

This gave a ~10% change in Q1 collector current. Probably acceptable?

better than no transistor.

 

[Flyback]

Thanks,
Incidentally The LEDs are power topleds by osram.

This alternative circuit (below) is even cheaper....and uses a vbe to regulate the led current to 50mA. (the 2.5vof the TL431 appears across the vbe and the emitter resistor)
I tried changing transistors to see the effects of transistors with different parameters in the different strings......in the simulator its not much difference...the simulator appears "kind" to this circuit.

The 2 Power BJT's in the actual circuit are BCX56-16
https://www.nxp.com/products/automoti.../BCX56-16.html

I really cannot see how accurate current regulation can be garnered with a BJT VBE......but this is happening inside cars today.

Would you agree that the VBE of the BJT = BCX56-16 could be anywhere between 400mV and 1050mV depending on temperature and batch?
BCX56-16 datasheet:
https://www.infineon.com/dgdl/bcx54...51dc7&fileId=db3a3043156fd573011589d20fda0368

if someone presented this to you as a automotive rear led light regulator, would you agree that the led current accuracy is insufficient?

LTspice sim and schematic of alternative attached

 

[audioguru]

What is all this "is cheaper" nonsense and the circuit with many parts?
Simply use an LM10 which is an adjustable voltage reference and an opamp both in one 8-pins package.

 

[MrAl]

Hello,

Whenever you decide to design a new circuit that is new enough such that there are no reference designs available or it is very different than most other designs of a similar application you need to do more analysis to make sure it works the way you think it does. For one, a sensitivity analysis for the various parts. Even if you dont do this mathematically you at least need to think about it.
For example, if the right transistor BE voltage rises then it draws less current, but then that means the other transistor draws more current.

Another viewpoint is that you should ask the question, "Is this circuit really any better in any way than any other circuit of the same type?".

We immediately see a problem issue about current sharing between the two LED strings with component variation. The two transistors wont conduct the same if they are somewhat different regardless how or when that difference came about to be. And when we compare this to a simpler circuit, we see that we have really just introduced problems into a circuit rather than come up with something that is in any way better.

For example, add a series resistor to the two LED string and maybe a smaller one in series with the three LED string and then drive them from the same transistor. Suddenly everything works and nothing to worry about.

From what it appears that you want to do, this kind of thing only works well at the die level. That's where the thermal coupling between transistors is much much better so they tend to track each other, and impurities are similar in the same die.

If you do want to stick with the original circuit, at least set it up so that the two emitter signals both contribute to the feedback signal (a couple more resistors) and that may help a lot.

There is one thing interesting about this circuit however. That is, what is the best way to control two linear transistor drivers with the same op amp error amplifier. We could look at this more too for what it is worth. For example, would it be better to use the same emitter resistor for both strings if the currents are equal. But the answer may be that it is better to drive one transistor with a higher current rating than to try to drive two.

 

[Flyback]

MrAl

There is one thing interesting about this circuit however. That is, what is the best way to control two linear transistor drivers with the same op amp error amplifier.

..you've hit the nail right on the head there...the whole point is to reduce component count and still get good regulation in each string....ie just regulate one led string, but make this one regulation ok to control 2 or more strings.
I thought the magic of "emitter degeneration" could do it but am not sure now.?

 

[MrAl]

Hi,

When you say "reduce component count" what exactly do you mean by that? Normally when we state that we mean that we have succeeded in subtracting from the overall component count in that one circuit alone, regardless what it takes to create other circuits that are similar but have to have different values for some of the components.
Maybe what you are referring to there is to "reduce component inventory stock" where we only have to stock say a couple component values for a circuit that will work with many different LED's for example. If we had several types of LEDs and needed 10 resistor values in order to accommodate 10 different LED types that would mean stocking 10 resistors for this one circuit, but if the transistors could regulate with any type of LED without the need for a resistor in each string then we might reduce our required stock to only 1 resistor value (at least for that one particular place in the circuit).

Assuming the latter, then we have something to dream about. But if you are going to dream then you darn well better know how to analyze the result, or know someone who will do the analysis for you. I will assume you know nodal analysis, at least for static DC circuits with dependent sources. Given that, you could replace the two transistors with controlled sources and test your hypothesis and see if your circuit works the way you hope it does. That will tell you if the circuit at least has a chance of filling in for the more conventional circuit. If it does, then you have something worth while. If it doesnt, then you have to dream up a way to get it to work, then analyze that. In the end you could actually end up with a very worthwhile circuit or at least you will then know the exact reason why it does not work well enough to be used for anything real.

If i get a chance and have the energy left today <chuckle> i'll take a closer look at this too and see what i can find out. But i know that you like to come up with new ideas like this and so it would be a good idea for you to do the analysis too. Then you'll be able to figure out which of your new circuits are worthwhile and which are not, and how to fix those that are not.

 

[MrAl]

Hello again,

I did a quick simulation too and found that the results are not too bad, and a little improvement is possible too.

With collector resistances different by 100 percent, the collector current was the same in both matched transistors:
0.0498, 0.0498

and with a 1 percent change in emitter base voltage of the first transistor the output current changed by roughly 1 percent.
With a 10 percent change in emitter base voltage the output currents changed as follows:
0.0498, 0.0534

which is not too bad really, and the first transistor has the same current as before.

Using two resistors each 10k to sum the two emitter voltages, the results came in a little better. The two collector currents with matched transistors were:
0.0497, 0.0498

which are still very nearly equal, and with a 1 percent change in Vbe again a 1 percent change in collector current resulted.
With a 10 percent change in Vbe the currents changed to:
0.0480, 0.0516

and here we see the first transistor current changed very little while the second transistor current changed less than without the two summing resistors.

In conclusion, the sensitivities with no extra summing resistors was about 0.72 percent and with the added summing resistors the percent change was about 0.36 percent which is half of that without the additional resistors. The non linearity of the base emitter diodes also works in our favor to decrease the sensitivities with increasing differential somewhat, in addition to the emitter feedback of each section.

So the circuit has hope.

What do you intend to do with this now?

 

[Flyback]

thanks MrAl, I am not sure how in LTspice you change the Vbe voltage of any of the NPN models......i suppose you just open the .sub file in notepad, find where it says "vbe=" and change it to whatever you want?

The circuit, as you say, appears to offer brilliant cross regulation of two (or more) strings of leds, even though only one string is actually directly regulated. There simply is no need for a separate opamp negative feedback loop around the emitter resistor of each led string.

Or am I discounting the effect of transistor heating, which ltspice doesn't do too well?...as described here, the transistor with only two leds will heat up much more than the one with three leds, and so its vbe will decrease significantly, and I reckon you could get one transistor with vbe = 0.5v, and the other with vbe = 0.9v

...this would give 44mA in one string and 56mA in the other string.....not too great. But is it so that the higher current would tend to occur in the regulated string (which only has two leds, so has a hotter transistor), and so at least the unregulated string would run *less* current than the regulated string, so therefore nothing would overheat so much?

I'd say this LED circuit is an excellent candidate for a rear/brake/indicator LED light on a car.....multiple strings could be used and they wouldn't even need regulation, because only one string needs regulation, the others are regulated by proxy, and also by the magic of the "emitter degeneration"?

I don't even think it matters too much if the pass transistors won't be well thermally coupled?

All those linear regulator IC's specifically for car rear led lights (on sale by semico's).....they seem to be a waste of money......as we see here, a couple of npn's and a single regulated string, and you're done......the regulated string isn't even that well regulated....its just 2.5V = Vbe + V(emitter resistor)...........and then I(led) = V(emitter resistor)/R(emitter)

beautifully simple, beautifully accurate, beautifully cheap.

Does anybody see a disadvantage?

 

[audioguru]

I have seen some LED rear lights on cars not working.
I have also seen MANY LEDs in traffic lights not working.
Is it the LEDs or the driver that messes up? Oh, I forgot about that awful solder with no lead in it.

 

[ronsimpson]

With this type of circuit, the current regulation is greatly related to the voltage across the current sense resistors. Now the voltage is 1V. If it was 10V (ridiculous) the Vbe effect would be 1/10 as much. If the voltage was 0.1V the regulation would not be good at all.

that awful solder with no lead in it.

At one place I work, you can have lunch at the bench with lead solder. You can not have lunch at the lead free benches.

 

[alec_t]

The disadvantage of relying on one regulator for all strings is lack of redundancy in the event of regulator failure.
I guess it's down to marketing whether you make cheap products which need frequent replacement (and lose reputation) or more expensive products which last longer (and boost reputation).

 

[Flyback]

With this type of circuit, the current regulation is greatly related to the voltage across the current sense resistors. Now the voltage is 1V. If it was 10V (ridiculous) the Vbe effect would be 1/10 as much. If the voltage was 0.1V the regulation would not be good at all.

..yes I see the point, and in my post #9 its actually 2.5V across [Vbe & emitter resistor], so about 1.8V across the emitter resistor, thus it is even more immune to vbe differences, than my schematic in post #1.

 

[Flyback]

Please could you advise on why the designer did not use an TL431 to provide closed loop negative feedback as follows...(please see below schematic "LINREG 1_TL431REG")
In the actual circuit, the designer , for some reason, just used the TL431 as a 2.5V zener diode ("LINREG 1_TL431REF").
Was this because using the TL431 in a negative feedback loop could engender oscillations?

Schematics and LTspice simulations of the circuit with the TL431 used as a zener diode, and as an opamp with negative feedback.
(please give me a shout if you wish to receive the .asy file and .sub file for the TL431)

Here are schems in jpeg
LINREG - TL431REF:
https://i42.tinypic.com/17qtsi.jpg
LINREG - TL431REG:
https://i44.tinypic.com/33dzitx.jpg