Why are car LED Rear/Brake lights done with linear regulators?

[treez]

Car Rear/Brakelights are only 4W each.
Why do people do them with linear regulators?
4W doesn't sound much but with the rear light being non-ventilated, that heat is trapped, and so can overheat.

Why can't an SMPS led driver be done as below.......?
This is a DCM Boost preceeded by a current clamp to prevent high current oscillations in the LC EMI filter if the driver "stutters" on and off repeatedly with the brake pedal.
From the moment the brake pedal is pressed, the output is in regulation just 1.3ms later......that ain't that much worse than a linear regulator....and it runs a whole lot cooler.

 

[betwixt]

I'm only guessing but brake lights are probably classed as 'safety critical' and would need redundancy. In other words, the LEDs should be driven from multiple sources in case one should break down. The cost of a linear regulator, duplicated several times is less than an SMPS duplicated several times.

Brian.

 

[dick_freebird]

Also, incremental 12V milliamps are roughly "free" so not
much incentive to chase efficiency at the expense of
up-front cost and long term reliability (numerous extra
parts). Besides, the LED probably draws less than the
bulb it replaces including the regulator waste current,
and the assembly is less dissipative than the bulb as a
result. Of course the bulb does tolerate high temps
better.

 

[treez]

Thanks, but a linear regulator is going to chuck out far more heat than a switchmode solution, and thus its electronics is far more likely to fail.

I say all indicators (which obviously double as warning flashers), rear lamps, and rear brake lights should be switch mode LED.
Why am I wrong?

Low power SMPS's are dead easy to design very quickly.
The worst enemy of LEDs is heat, so why not switch mode led drivers for all rear lamp requirements on cars, ie, brakes, flashers, rears etc etc?

 

[betwixt]

I would argue that the worst enemy is moisture, not heat. Semiconductors, including LEDs tend to live longer when generating heat as it prevents moisture ingress along the connections. Obviously there is a point at which the heat becomes excessive but running them cold wouldn't necessarily mean they last longer.

As has been mentioned in your previous posts, SMPS suffer from capacitor problems, especially after prolonged periods, for a safety critical part of a vehicle which needs 10+ years of reliable operation, linear is a good option. Remember that efficiency isn't so much of a problem when you have lead acid or Polymer battery and a generator running most of the time. There is also the issue of multi-LED strings needing close to 12V anyway so a regulator would have a fairly easy life.

Brian.

 

[treez]

Regarding caps...yes but smps for rear lights can be done with ceramic only, alleviating those problems.....there's still some heat generated with smps, so its not running totally cold.
I find you cant count on the led string being 12V.
At cold ambient, the leds have much higher Vf.
So, realistically, when the reverse polarity diode, and the wiring drop is considered, you have about 11.5V to play with.
Four red leds in series, in cold weather, can be 12.4V......so you cant use 4 per string...so youre stuck with three, and at high ambients, three red leds can be just 7.2V..meaning that you are wasting (11.5-7.2) * current in power.

For a 4W rear lamp, you are going to be wasting 4W dissipation, and that is going to overheat the leds in places where its hot weather.

 

[betwixt]

I'm not sure why a reverse polarity diode is needed when there is no risk of misconnection, it one was needed to protect against negative spikes, it would be across the supply and hence not dropping any voltage. Typical voltage to a tail lamp cluster is about 13.5V with the engine running and about 12V when it is stopped so chaining 3 LEDs at say 200mA would require a regulator to drop around 13.5 - (3 x 3.1V) = 4.2V and dissipate 4.2 x 0.2 = 0.84W, possibly some of that in a series resistor. I don't think that would be a problem. From a costing angle, in bulk, SMD 9V 1A regulators cost about 0.20 (sterling) so using several to build in redundancy is very cost effective. I can't see SMPS ever being as cheap or as small as that. There is also no need to accurately regulate the current, even under cranking conditions when the voltage may be as low as 5V there is no need to use a boost technology, conventional lamps just dim down and so would LEDs.

Brian.

 

[treez]

well, below is the rear lamp for a Bentley...with multiple linear regulators.
The LED power is 5.94w assuming 3V per led.
When its hot, and the led vf is about 2.4v, then led power is 4.8w.

total current to led banks is 12*0.055 = 0.66A.
The "wasted" volts is 13.5-7.2= 6.3V
.....So the wasted power is 6.3*0.66 = 4.158 watts, and that's too much in my book, is it not in yours?

i confess i don't know if the 4.8w is "braking", or just rear light...looking at the schem, it doesn't look able to vary the current between "rear" and "brake" levels.....im pretty sure that "brake" lights are at least three times the power of "rear" lights(?).

one thing I do remember, was that for light level testing , it had to be left on for 30 mins minimum before they did the test...and the design engineer was having kittens because all those sot23 bjt's etc were getting well hot...he was scared the whole assembly was going to melt.

 

[betwixt]

The wasted power is only about one quarter of the amount lost as heat in a conventional lamp so I would consider it a better design. I appreciate what you are saying though but wonder if you could produce less heat from 12 SMPS units (losses ~300mW each) and at the same time guarantee 10 years minimum lifetime.

I wouldn't lose too much sleep over it getting hot, in a previous job I did, we took complete PCs (for a VERY large International Computer Company) and baked them for 5 hours at 70C then took them down to 0C for 5 hours and repeated the cycle several times, making sure they passed all their diagnostic tests while continuously being powered up and down again at both minimum and maximum line voltages. They were done in batches, around 500 at a time and the power to them was rated at around 40 KA.

I'm a bit puzzed as to how the diodes work in the schematic, I think it reads "tail monitor" but they are configured more for injecting current than sensing it. Perhaps they are part of the tail/brake power control.

Brian.

 

[treez]

it does read "tail monitor".
I think they are for sensing when a string goes open.
I wonder tho' as you suggest, they may be for injecting current as part of dimming down to "rear lamp" light level(?)

 

[treez]

If you look at the rear lamp schem in post #8, you can see how bad it is...the transistors dissipate (in hottest ambient) above 150mW....so these cannot be SOT23. They must be at least the next size up which is SOT89. I am sure you will agree that 12 pieces of SOT89 is a poor solution, taking up much space, and really, an SMPS solution should be sought after.

These LED open protectors can provide protection against open circuit failure of any LED.
https://www.onsemi.com/pub_link/Collateral/NUD4700-D.PDF

 

[schmitt trigger]

As betwixt has mentioned, I also believe that redundancy is mandated in a brake light, and as such, it is easier/cheaper to use multiple linear regulators.

I have seen a few vehicles where a couple of strings have gone defective, nevertheless the remaining string do light up.

 

[treez]

in that case, I would use two smps, if the above led open protectors aren't already enough.
Having a hot linear regulator type internal ambient, will increase the chance of total failure.

Would you really use twelve SOT89's in the schematic of #8 post?

 

[schmitt trigger]

I have done my fair share of automotive design reviews. Different companies have different approaches, but most use a variation of the "firing-squad" method.

Essentially, one creates a preliminary design, and then creates an extensive power point presentation, outlining the design criteria, costs, all the pros/cons.
Then one then goes to the customer's corporate design center, and one is ushered into a conference room with anywhere from 10 to 20 people eagerly waiting for you.
Those people sitting there will cover the whole spectrum of mechanical/electrical/software engineers, accounting, quality, manufacturing, purchasing, styling, and all sorts of assorted "experts". Reliability experts, thermal management experts, EMI experts, lighting experts... you name it.

One delivers the presentation, and in the darkened room one can see the glint in the eyes of the persons intently watching you.

One finishes the presentation and then naively ask: "any questions?"...of course there will be questions, tons of them. And comments, lots of criticism, the occasional congratulation.

Depending on the complexity of the project, the ordeal may last a couple of hours or a couple of days. Certain things are a sure deal...your ego, if you had one, is turned to dust. And your design, shredded to pieces. And dozens of action items for your next meeting.

Actually, if one is humble enough, it is quite a learning experience. The people assembled in the conference room represent an extremely wide array of knowledge and experience, and they may have totally different insights, be aware of certain trends, understand certain regulations, or know a few obscure failure modes that one was oblivious to.
For instance, you ask: would you use all those SOT89 transistors? It could be that the automotive company has contracted to purchase a gadzillion/year transistors from XYZ semiconductors. Their cost, even after assembly, could be lower than sawdust.

Before this Christmas vacation, we had one such a meeting. Our customer, the world's market leader in heavy machinery, has found out a failure mode straight out of science fiction. This will radically impact our product design and validation.
Regretfully, I cannot disclose details.

 

[betwixt]

Been there and done that. I remember going throught the process for a prestigeous car manufacturer with orders from rich Middle Eastern customers. The design was quite complex, especially from the mechanical aspect. when I asked what quantities they were looking for they said "five". I queried if this was five thousand or five million, the answer was just five!

Going back to the redundancy problem, I doubt whether two SMPS would be regarded as acceptable. To be 'fail safe' I would think they would be looking for at least four parallel circuits and as SMPS would cost significantly more, I don't think they would see it as a viable method. The other issue which has not yet been discussed is whether the dissipation in the SOT23 is actually 150mW since no method of controlling the LED brightness has been mentioned. It is quite possible the control signals are PWM and therefore the heat produced would be proportional to the drive waveform and would only be 150mW at constant 100% 'on'. The TL431 can react very quickly in that constant current arrangement but whether it is passing current or turned off would depend on the control signal. Possibly, that explains all those capacitors across the transistors.

I have a LED lighting system here which runs on 12V, dumps around 2A to an LED array and is driven from a single TO-92 MOSFET. It can be controlled to 1024 levels and the MOSFET runs cold at all of them. It can be done.

Brian.

 

[andre_luis]

As said above, decision to select Linear Regulator topology instead SMPS, may me merely casual.
However, there are some reason which could contribute to avoid selection of circuits based on switching.

If we consider that - differently from LAMP devices ( which are able to convert energy to heat ), LEDs doesn´t support operate reliably outside SOA region, and could prevent failure be not powered by switched circuits which are able to produce over rated voltages, even for a small period.

A thread related to this subject was raised here, take a look: https://www.edaboard.com/threads/220363/#post937189



+++

 

[treez]

It is quite possible the control signals are PWM and therefore the heat produced would be proportional to the drive waveform and would only be 150mW at constant 100% 'on'.

At high ambient (referring to schem of post #8)
Leds voltage = 7.5V
Volt drop in sense resistor = 1.2V
Bus voltage = 13.5v
therefore BJT (vce) = 13.5 - 8.7 = 4.8v
therefore bjt power = 4.8*0.055 = 264mW

264mW , I am sure you will agree, is way too much for a sot23 bjt.

..and that's even if it is only during 100% on pwm dimming. (ie when brake lights are on)

....I don't buy the "brake lights are only on for a fraction of the time argument".....if you are in a long downhill traffic jam, many people will keep the brakes almost permanently on, only releasing them for seconds at a time every 2 minutes or so , so as to let the car roll forward.

Surely you will agree, 100mW is absolute max for a sot23 BJT?

I see the argument about gadzillions of sot 89s......but one could say that about any component.....and it would be folly to always do linear regulators based on "they might be buying gadzillions etc etc"

I say for 5W of a brake light.....just do an smps....the design of such would take about 4 hours to break the back of the design....no more than that.

If they really needed 12 separate strings for redundancy, then i would use 12 of these zxld1360 based buck converters....
**broken link removed**

.....twelve half Watt buck converters side by side on a pcb will not be noisy....certainly no noisier than a single 6W switch mode converter....and even with the twelve (small) inductors, it will be less space taken than 12 SOT89 BJTs used in linear mode.

switched circuits which are able to produce over rated voltages, even for a small period

thanks, though I am sure you will agree, its easy to make smps led drivers which don't produce overvoltages on the leds.

 

[betwixt]

ZXLD1360ET5TA devices would be a good choice from an electronic point of view but are far more expensive than single transistors and a 431 regulator. Most SOT23 transistors will happily dissipate 250mW while soldered to FR4 substrate, I would guess that in a cluster light, the substrate would be metalic so they could be well inside safe dissipation limits. Don't get me wrong, I'm certainly not 'anti SMPS' or anything like that but many years of experience with customers teaches me that their design brief goes like this "1. the budget is <a small number>, 2. it has to do this... 3. there is no 3.". Money talks, engineers rarely get a say unless they can find a way of cutting costs. The other point I would add is that you shouldn't assume that the original designer was an expert or even had a clue what they were doing. I have seen college students on 'work experience' release set the task of designing a product, many never having designed anything before or had any manufacturing tuition. The result is invariably something cobbled together from various application notes on data sheets with no regard for how they relate to each other. For example:

(he's dead now so I can safely quote this)
An audio test oscillator. It had a single transistor and a simple RC phase shift network. The data sheet said the design procedure was to select appropriate capacitor values and calculate the resistors needed. Being inventive they did it the other way around, they picked 100 Ohm resistors and worked out the capacitor values, I forget the exact value but lets say it was 1.2345uF. They then contacted a capacitor manufacturer and ordered them as 600V polypropylene types with 1% tolerance. They ordered five and each cost > £100 to be custom made. It was a very expensive and very large oscillator and didn't even work particularly well.

Another instance, same company, different 'engineer': Built a logic circuit for driving a row of 7-segment LEDs. They correctly realized that their CMOS ICs couldn't sink or source enough current for the segments so they used all of the 6 inverters in a CD4049 in parallel FOR EACH segment. Because the signals were inverted from what they wanted, they pre-inverted them using ULN2803 high current drivers! The result was a board with about 60 ICs on it, all inappropriately used when the job could have been done in two.

I could tell you the story about the students designing a guided weapon but the MOD would shoot me if I did.

Brian.

 

[treez]

Most SOT23 transistors will happily dissipate 250mW while soldered to FR4 substrate

the following 2N3904 datasheet states 357 degC per Watt for SOT23 and that's with 1.6" square PCB copper on top and bottom......so 250mW will raise it 90 degrees....on minimum footprint, it will obviously be much more than that......I would be concerned about 250mW in a SOT23...youll need a big PCB which in itself will increase costs.

I would guess that in a cluster light, the substrate would be metalic

...a metallic substrate would be really expensive, I am sure youll agree that smps on FR4 is cheaper++ than linear_reg_on_metal_substrate.

 

[schmitt trigger]

Have you ever been personally involved in a "firing squad" design review?
By personally involved, I mean that you are the one standing up, receiving all the flak.