LEDs turning off causes EMC problem?

[treez]

hello,

i am doing a lamp with 8 300mA LEDs powered by SMPS.
Source is 13.5V

The LEDs must come on one-by-one in a line then stay ON.

The simple way is to put the 8 in series then switch out FETs which short out each of the LEDs.

However, this means that just as the next LED is switched ON, -all the LEDs (that are ON) go off then on again.
(since initially there is not enough forward voltage to light the LED string with an extra LED in it -until the voltage builds up).

.......this cannot be seen as they only go off for 1ms but the sudden OFF-going of LED current will cause a bad EMC problem.......is this true?

...how may this be mitigated?

 

[VEBill]

The shortest possible answer is to use an RC on the gates of the FETs to slow down the transitions so that your PS can keep up. And there'd be no concern about EMI from such slow transitions. The heat from the FETs operating in the linear region can be controlled by the rarity of the cycle.

But there's probably a much better circuit architecture than the one you've described. The approach of shorting out LEDs is very odd.

 

[IRstuff]

"-all the LEDs (that are ON) go off then on again"

That's something unique to your design. It doesn't have to be that way.

TTFN

FAQ731-376

 

[MacGyverS2000]

It has been said several times before, but it's starting to become obvious... you need to bring in outside help with this one. We've practically designed 50% of the system for you at this point, and you're many, many months into the design with no end in sight. This is beyond simple suggestions, it's actual repair work. Sometimes it's okay to admit you don't know how to solve a problem...

Dan - Owner

http://www.Hi-TecDesigns.com

 

[treez]

macgyvers2000

"It has been said several times before, but it's starting to become obvious... you need to bring in outside help with this one"

"Sometimes it's okay to admit you don't know how to solve a problem... "

I am afraid my boss has been trying to recruit a switch mode designer for 18 months with no success at any wage offering.

The team of 5 who are here have managed to get by so far (apart from the 65mm by 15mm double sided PCB 10W LED driver which we haven't been able to do for the last eighteen months as it fails EMC.)

 

[zappedagain]

Your on-off transients are causing voltage/current surges on the lines. That is a classic source of EMC problems, although at 1mS yours may not be a problem. I suggest you get a copy of

EDN's Designer's Guide to EMC

from Kimmel & Gerke for $29.00 (

http://www.emiguru.com/books.htm).

This is an excellent reference and will get you started on figuring out what is and what is not a potential EMC/EMI problem.

John D

 

[treez]

thankyou all,

by the way, we're going to go for this LED chaser...

heres the output side...

heres the input side

........
its only easy electronics so my boss wont mind me posting.

any improvements or ICs for the output current limiter welcome.

(we have to limit output current as it may go on-off-on and the top led would suddenly get 15+ volts across it.)

we are going to limit off-going of leds sharp-fall-ness by RC'ing the fets which switch in the leds progressively.

 

[OperaHouse]

That sure is a lot of parts. Can't even imagine how many parts are used in the sequencer. Shorting out LED seems like the last choice for a sequencing circuit. Are you trying to use existing lighting hardware?

 

[treez]

Operahouse:
thankyou for your interest.

You are quite right, it is company policy to re-use as much of already existing circuitry as possible.

Having said that, it would be difficult to do this with less parts-count than the above.

This is an alternative but needs eight inductors and eight EMC filters (one for each LED)

The LED chaser only happens at start-up and only once. - so there are few other components than the above other than the PIC.

Any ideas for a low parts count method though greatly appreciated.

We dont have top SMPS grade designers here...there is simply not enough work here to keep such people occupied.
-they dont want to come here at any price....but my boss is still trying to get an expert in -unsuccessfully.

 

[treez]

Coming back to my earlier point if I may….,

The LEDs of our lamp come on one by one then they all stay on.

This means that as the next in the sequence is brought on, all the LEDs current suddenly go to zero (for about 1ms) , then rise back up again as the LEDs’ forward voltages builds up to a sufficient level.

The point is, would this sudden zero going of current create an EMC problem?

I wouldn’t have thought so because this event only happens for a miniscule portion of time just after the lamp is switched on…..so I would have thought it will not matter as EMC readings are averages taken over significant time periods ?

 

[treez]

sorry, i forgot this diagram showing the LEDs coming on in sequence, and the way the LED current snaps down, just before the next LED is brought ON.

(the FETs that switch in the LEDs, have had RC (10K, 1uF) filters put on their gates , to delay their turn-off, but it doesnt help that much as you can see

 

[mtripoli]

I'm jumping in here late, but why are you doing this this way? There are many, many much easier ways of doing this. An led driver IC and a microcontroller (after your SMPS) is a much easier way to go. Check out the TI part TLC5917 for an led driver. An offline switcher like the "Viper" series from ST has less parts count than your LT part I think. Any Microchip PIC or Atmel AVR can talk to the led driver. You can PWM the leds on and off and avoid any of the EMC issues you are seeing.

 

[treez]

Hello mtripoli,

Thankyou,

But we cant use TLC5917, -only 120mA and Vin too low
Also, Viper switching freq is too low.

This way we do it is lowest componen count way possible –or so we believe.

We have a pic to switch out the LED-shorting-FETs.

We cannot do PWM of each LED because that needs 8 LEDs in parallel which is too many components.

 

[MacGyverS2000]

We cannot do PWM of each LED because that needs 8 LEDs in parallel which is too many components.

I do not understand this statement. I routinely PWM 8+ high-power LEDs directly using a 16-pin PIC16Fxxx, all the while communicating to it over an RS-485 link. Other than the 485 driver and resistors/caps, I have little more than one FET per LED in the system.

Your engineering team has over-engineered this...

Dan - Owner

http://www.Hi-TecDesigns.com

 

[MacGyverS2000]

In case it wasn't clear... I was talking about each LED having its own PWM pulse train, and Knight-Rider style displays would take no more than 5 minutes of programming a sequencing function.

You guys need to go back to the drawing board and decide what you really want out of this thing.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[mtripoli]

I agree with Dan; this is a very simple project that has gotten way out of hand. I have never in my life heard of turning things off by "shorting them out". Goofy question maybe, but did you guys start with a "specification"?

 

[OperaHouse]

It seems you were havimg noise issues with the SMPS in another post without any LEDs, so the LED drive is not an issue. Until you can define the noise problem there can't be a solution. I worked with a mechanical engineer who lacked a certain regard for our electrical profession. He seemed to think that any time we had a problem with a circuit we just put more capacitors on it. If I had to guess, you are operating the inverter at a really high switch rate to reduce the size of the inductor. That can buy a lot of problems for a neophyte designer as component selection and layout are critical. At a slower rate there probably wouldn't have been issues.

Problems are technical and solutions are political. A top down redesign is likely unacceptable to those in charge. A PIC could easily perform the SMPS function. The saved components would leave room for a larger inductor. Tweaking timing could easily resolve any timing glitch issues. Buying a separate SMPS chip just wastes the power of a PIC. A PIC casn easily monitor LED current and best of all it knows when current changes will happen. Think of this as a total system and don't show that posted circuit at your next job interview.

 

[treez]

ok Thanks, we are operating at 500KHz as usual for us.

The spec for this converter is in the top few lines of my first post.

It’s eight 300mA LEDs.
They don’t all come on at once, but come on one-by-one in a line, thenm just stay on.

I know nobody suggested it, but we cant do any linear current regulators because the power supply and LEDs will be in the same sealed enclosure, possibly in the burning Arizona sunshine and just maybe the customer is going to have lights ON in the middle of day even if by accident.

Our input voltage is 13.5V, but has transients down to 8V and up to about 17V.

Since we need the LEDs to come on one-by-one and not all at once, this has created the unavoidable complexity of my above circuit.

I am sure there is no easier way but I have the greatest respect for you if you have found an easier way and would be grateful for some clues as to its features.

The only other way to do this is to have eight switch mode current regulators supplying each of the LEDs, and then switching them on in sequence…..

…but eight switch mode current regulators, -even though they are only around 1 Watt each, means 8 inductors, eight Schottky’s, eight smoothing caps….etc etc. and then they need synchronization, and then there’s the LED failure detection circuitry which will add up for eight such separate supplies –not forgetting the problem of switchers giving output overvoltage when a LED fails open.

-I admit that we could use two LT3476’s (with on-chip FETs), which are 4-channel LED drivers, this would give eight LEDs in parallel, but that would still mean eight inductors and Schottky’s.

Also, the two LT3476’s could not be synchronized with each other and would beat with each other.

I actually like the idea of using two LT3507 triple channel LED drivers and one LT3508 dual channel LED driver (all with on-chip FETs) to get eight LEDs in parallel, but have been told I’m a nutshell for even contemplating a solution with eight inductors.
(LT3508 AND LT3507 are really PWM controllers but could be adapted for use as LED drivers)

Sorry, I hope you can prove me wrong, in fact I’d like it very much.

But there appears to be no easy way round this one…..and all because the LEDs must be controlled to come on one-by-one –and not come on all at once.

(By the way, thankyou to VE1BLL for your idea of RC’ing the FET gates, -it works well and smooths out the on-coming of the next LED.
I used 10K and 1uF as more would have meant it taking too long….current still dips with these values but I’m hoping its ok EMC wise.

 

[MacGyverS2000]

Problems are technical and solutions are political. A top down redesign is likely unacceptable to those in charge.

A team of 5+ engineers has been working on this for 6+ months, and you think a top-down redesign would be unacceptable?

Dan - Owner

http://www.Hi-TecDesigns.com

 

[KnoItAll]

Some ideas...
1. put some inductance in series with the LEDs
2. use your voltage source to drive a current source, then connect the LEDs.

http://en.wikipedia.org/wiki/Current_source