Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

LEDs in parallel is OK?

Status
Not open for further replies.

grizedale

Advanced Member level 3
Joined
Jun 13, 2011
Messages
838
Helped
17
Reputation
34
Reaction score
17
Trophy points
1,298
Activity points
8,804
Hello,My company has been putting 1A power LEDs from Cree and Osram in parallel for soome time now.

We only ever use a single current source, and NEVER use series current equalisation resistors or current mirrors.

We have encountered no problems......so why is everyone telling us we are doing wrong?...i just dont understand. A batch of the same type of power LEDs always have virtually exactly same Vf's, and so no LED hogs the current.Driving paralleled LEDs with no equalisation resistors is the cheapest way to drive LEDs.....you can just usually use a cheap simple hysteretic buck converter......Why are so many people (even some prospective customers) telling us we are doing wrong?....we have had no problems. On our production line, each finished product gets left ON for 20 minutes before being bagged and despatched..........if a failure was going to occur, it would occur there...but it never has, so we are thinking of doing away with this mini soak test. Our LED lights are in the power range of 2W to 50W.

Why is what we are doing wrong?
Do you agree with us that this talk of unequal Vf's is a myth?
 

I think you are pitting theory vs. practice. If your Vf's don't differ that much, then you are probably ok. Because these are LEDs and not IDEAL diodes, you can operate on the rounded part of the V-I curve. You might see differences in light output from one LED to another, but the fact is, the specification for the LEDs is pretty broad anyway (usually about a 2:1 range) as far as light output for a given operating condition.

Of course, there are people out there who also believe that gold-plated speaker wire is the only way to go because of phase-shift or some other nonsense.
 

A popular chain in the UK (Maplin) tried selling me a optical audio Hi-Fi cable with a gold plated plug, because the gold plating means it is higher quality.
**broken link removed**

Are you suggesting they were lying?!!!
 

Hello,My company has been putting 1A power LEDs from Cree and Osram in parallel for soome time now.

We only ever use a single current source, and NEVER use series current equalisation resistors or current mirrors.

We have encountered no problems......so why is everyone telling us we are doing wrong?...i just dont understand. A batch of the same type of power LEDs always have virtually exactly same Vf's, and so no LED hogs the current.Driving paralleled LEDs with no equalisation resistors is the cheapest way to drive LEDs.....you can just usually use a cheap simple hysteretic buck converter......Why are so many people (even some prospective customers) telling us we are doing wrong?....we have had no problems. On our production line, each finished product gets left ON for 20 minutes before being bagged and despatched..........if a failure was going to occur, it would occur there...but it never has, so we are thinking of doing away with this mini soak test. Our LED lights are in the power range of 2W to 50W.

Why is what we are doing wrong?
Do you agree with us that this talk of unequal Vf's is a myth?

No my friend you doing wrong of course, you will not kill led imidietly but you will degrade led illumination. Led without resistor can drain more current then can handle. Also temperature will increase, that specially is bad for led. You dont use resistor but maybe your equipment have implemented current limiter inside. Each manufacturer give instructions for using their products, some even provide and dedicated power supply for leds. Paralleling different leds with different characteristics can cause that one led suffer. I dont think that there is some manufacturer of led which will suggest to use leds without resistor or other current limiting technique. Sorry my friend I think your company is ready for some inovations.

**broken link removed**
 
Last edited:

No, not at all. Gold is definitely better for contacts because it doesn't oxidize like other metals will. My point was that there are supposed 'audiophile' cables that have "lower phase shift" than conventional (i.e., not stupidly expensive) ones. Nobody, NOBODY, can hear the difference. It's the same for audio amplifiers. 99.9% of people can't detect distortion below, say, 0.1%, yet there are amplifiers with 0.001% for people with extremely well-developed bank accounts.
 
@Barry, hehe yes I agree, I just thought it was rather funny that the store were trying to sell the same snake-oil with
fiber-optic cables with gold-plated connectors : ) Apparently it's ideal for "advanced acoustic equipment" ; ) Someone
should report the manufacturer to Trading Standards : )

Pure plastic optical fibre (POF) clad in special reflecting layer to prevent phase error
Gold-plated TosLink™ connectors
Includes special mini-plug adaptors
Ideal for advanced acoustic equipment
 
We have encountered no problems......so why is everyone telling us we are doing wrong?...i just dont understand. A batch of the same type of power LEDs always have virtually exactly same Vf's, and so no LED hogs the current.
This is not always the case. It all depends on the type of LED, and the quality of the manufacturer. Apparently most people don't trust the manufacturer enough to leave current balancing in their hands.
each finished product gets left ON for 20 minutes before being bagged and despatched..........if a failure was going to occur, it would occur there
Probably not. When stressed, LEDs don't always burn out immediately. It can take many, many hours. Especially when you consider that current imbalances are increased by positive temperature feedback.
Do you agree with us that this talk of unequal Vf's is a myth?
No of course it's not a "myth." There will always be some degree of mismatch. Whether the mismatch is tolerable and whether balancing resistors are necessary depends greatly on the requirements of the design and the LED itself. Balancing becomes quite critical as you push the LEDs towards their maximum current ratings.
 
The problem will appear long term. The poor current sharing will cause some LEDs to be overloaded and therefore be too hot. The degradation of these and the chances of premature failure due to overheating are great. If one LED fails it will cause other LEDs to fail prematurely.

I would only parallel LEDs if system cost objective was absolutely critical (If you use Cree and OSRAM LEDs I understand that cost is not the only design priority). In this case I would make sure that the LEDs used are not only the same model but the same manufacturing BIN.

I worked some years in a LED lighting company. Seriously, if your company pretends to offer good quality LED products with the expected lamp life you should consider stopping your production and redesigning your system.
 
Though Cree LEDs are just as cheap as any other?

for example, Cree's XPEWHT, 1A White LED is the cheapest LED of its type?

Who is doing LEDs cheaper than Cree and Osram?
 

LED in parallel will work for months or years without any noticeable failure.
It will have a shorter lifespan than a properly design LED circuit.

Hello,My company has been putting 1A power LEDs from Cree and Osram in parallel for soome time now.

We only ever use a single current source, and NEVER use series current equalisation resistors or current mirrors.

We have encountered no problems......so why is everyone telling us we are doing wrong?...i just dont understand. A batch of the same type of power LEDs always have virtually exactly same Vf's, and so no LED hogs the current.Driving paralleled LEDs with no equalisation resistors is the cheapest way to drive LEDs.....you can just usually use a cheap simple hysteretic buck converter......Why are so many people (even some prospective customers) telling us we are doing wrong?....we have had no problems. On our production line, each finished product gets left ON for 20 minutes before being bagged and despatched..........if a failure was going to occur, it would occur there...but it never has, so we are thinking of doing away with this mini soak test. Our LED lights are in the power range of 2W to 50W.

Why is what we are doing wrong?
Do you agree with us that this talk of unequal Vf's is a myth?
 

If led receive higher voltage or current, brignest will be lower very soon.

Yes yes. Agree. I understand.
In practise, it may not be so soon.
Depends on the LED quality as well.

I am using LED lamps at home.
Out of curious, I study their design.
The LEDs are connected in parallel of 3 then in series.
I don't find it a good design; It works quite ok so far.
Practical and theory is a different story.
We should still follow the good design practice.

Best Regards,
Siong Boon
 

I agree. This can be different from case to case, and lots depends from quality of leds. In many cases manufacturer lying for leds full brightness life period, after that leds losing brightness. On that if we add more current and overvoltage I think this will make things worse.

Lets keep good and proper design.
 

does any person know of a LED product that DOES use LEDs in parallel , without any equalisation resistors or current mirrors?

i feel cartain that our competitors are getting away with it.
 

I've worked in Electronics for over 35 yrs in R&D, Test Engineering and Manufacturing. So Component Engineering is one of my expertise. LED's are a specialty in the last 5 yrs. In my spare time during retirement, I distribute ~1m LEDs per year and all are custom spec'd by myself to meet or exceed customer's requirements. One thing they do not have to worry about is Vf consistency for parallel operation.

LED's have a lower ESR than most zeners at the voltage and that is the over-riding factor in determining if the LED's in parallel are compatible, since the threshold voltage is fairly consistent. From my experience Vf variance can be as little as 0.1% in a batch. This is due to consistency improved in LED foundaries for wafers and sorting for LED's. YOur experience may vary for supplier to design application, but none of my clients have any issues from current hogging in parallel operation assuming you can allow 10% variation and not running it near Absolute Maximum Limits for current or junction temperature, otherwise diode voltage-temp effects can spoil the simplicity of your design.

ESR is everything in LED power for efficacy and maximum output.


- A 70'C rise in Tj can drop the voltage 5%
- Vf deviation on LED's in a batch tend to be <<1% and often < 0.1%
- Parts binned in the same batch { say of 0.2V range } are all exactly the same
- Parts supplied from 2 adjacent bins for Vf will not be exactly the same.
- if the design is conservative with margin for Tj or Imax then many designs can tolerate parallel operation.
- if the design is not conservative, you may have current sharing issues
- Generally designers who don't understand choose large values of resistance for efficiency is much worse
- if you understand the sensitivity parametrics, design limits and ESR using a large population you can estimate adding a value = ESR of LED
- This results in significant reduction of variance of Max/min current ratio.
- If say ESR tolerance is 20% of nominal adding 100% of ESR in series reduces the variation to approx. 10% Max/min current.
- If you do the math, you may determine if your design with your supplier needs ESR equalization.
- Again this questioner indicates they have no issues in production from one batch of parts. My customers have the same experience.
- ESR drops as chip size and current ratings increase
- When LED efficacy increases the result is usually a drop in ESR.
- e.g. 5mm LED's are ≈10 Ω for ESR at rated current, assuming High Bright White(HB) (75mW max)
- e.g. . 1W LEDs are ≈ 50mΩ , assuming 320mA * 3.3V
- My rule of thumb for diodes of different power ratings P1,P2 and ESR of R1,R2 their products are as follows;
- R1 * (P1)² = R2 * (P2)² . . .
- using 5mm HB parts 10 * (0.075)² = 0.056
- using 1W parts in white 1.0 * 0.050 = 0.050 shows good correlation for ratiometric estimates of ESR in same technology.
- testing the incremental ΔV/ΔI = ESR at design operating level is most relevant in any case.

- All LED's can be modelled by linear regression with a threshold voltage Vt and an ESR value to get the Vf at rated current. ( use >=3 points )
- This is useful for predicting normal operation.

When LED's fail, most often it is open circuit. So when one fails the rest of the lump share the current so conservative design can handle this situation.
Often smart clients will choose a suitable matrix of small series strings in parallel such as 12V or 24V or whatever suits their market.

What about SERIES LED STRINGS?
-
A far worse problem is PWM on a series string of LEDs.
- If one fails they all fail !!.
- Open circuit failure modes often get triggered by thermal and mechanical stress on gold wire joint at high temperatures.
- Steady state DC is ok but some designs shut down abruptly faster than the reverse 1MΩ typ leakage and junction xxx pF values.
- When pulsing LEDs the junction capacitance changes with voltage and off state and reverse voltage leakage resistance can vary significantly
- This RC parallel equivalent circuit forms a value in series with all the others. If they are all equal. The voltage declines equal. If unequal then the device with the highest resistance gets REVERSE BIASED.
- This can occur when the voltage is shut OFF, or when PWM'd for a large string with dimming control. Excess Reverse Voltage is catastrophic to LEDs and the entire series string. Solution have a bypass diode that conducts should the Vf be exceeded from a blown part.
- I have witnessed this failure mode and discovered the root cause at one client's site. It exacerbates other failure modes with this design flaw, when other Over-Stress conditions exist in the process such as ESD from plastic encapsulation.

Some solutions include..
- limit the rise time on PWM of LED connected Series strings
- better - use back to back LED's if driving with low current AC current or bipolar current source
- for use back -back Zener protection diodes to prevent highest reverse voltage on the device with the highest impedance and conducts if device fails with conservative current in series.
- use series values of resistance in the range of ESR for conservative parallel design
- use parallel leakage resistors to protect non-zener LEDs from rapid shutdown pulse.

- BEST: use back-to back Zener protection inside LED's at marginal cost increase.


Any questions?
 
Last edited:

Thanks sunnyskyguy

Vf variance can be as little as 0.1% in a batch

"can be"............if its only "can be", then i think ill forget paralleling them.

Parts supplied from 2 adjacent bins for Vf will not be exactly the same.

.....so are you saying that we have to go to the extra expense of buying binned LEDs so that we can do parallel operation?.......already that puts me off parallel operation.

Vf variance can be as little as 0.1% in a batch

..we get our circuits populated elsewhere, how do we know that they are using only leds from "one batch"............we find that we get lots of mistakes from our electronics manufacturers, eg misplaced components etc etc.........i dont know if i could trust them to only use leds from one batch.

you speak kindly of allowing 10% variation, but allowing any variation will lead to more variation, because of the thermal runaway effect.

We cannot afford specialist suppliers......we have to use the cheapest around, we dont get individual custom service from them, and we cannot afford to pay for that......so i am not sure that we can do parallel leds now, as we just can't afford the expaense of the extra work to ensure the leds are all batched and matched, so to speak.
 

then I suggest you do a statistical analysis on ESR and Vf and characterize samples from different batches. Then determine if the in-circuit ESR with the LED ESR variation and driver margin to limits are sufficient. I am a specialist supplier but my overhead is less judging by the markup of other Disti's. If you would like my assistance on the analysis, just ask.

Hope this helps.
 

Barry , . 5mm LED's in HB White (>30Cd @30deg) have a much lower ESR than even NXP's Zeners of a similar size.
That is even more significant with higher power LEDs vs Z's. .... knee is ~5% of rated.

Take NXP for example, one of the best suppliers

p/n PZU3.3 ESR = 95 ohm @ 5mA
typ. HB LED ESR= 20 ohm@ 5mA

Grizedale, Normally my clients simply what they need/want/don't want, then I translate that into a part number and spec. They any EMS or CM can buy them. I always get better parts for same or less price because I never carry inventory and ship build to order in 2~3 weeks. Just so you don't make a wrong assumption.

Let me know what you end up doing.
 
Last edited:

OK Thanks Sunnyskyguy, its certainly interesting to know that distributors such as yourself exist with that service.

Of course the other point is that LEDs in parallel, even if having exact same Vf, need to be well thermally coupled or else there Vf's will vary and one will go into thermal runaway.

I know that Tridonic is a huge LED lighting maker, and they do not use parallel leds.

.....however thats not strictly true though , since they have a special LED module built for them by the LED foundry which comprises 6 matched LEDs, being matched at the LED foundry

..and these LEDs are mounted on to a well thermally coupled PCB at the LED foundry.

Otherwise at Tridonic, parallel LEDs is a no no, and indeed is a taboo subject.


By conservative design i take it you mean de-rating, and using LEDs of higher current rating than needed....again thats more expensive..........and using heatsinks bigger than needed is more expensive.

I have often asked LED makers to give official documentation saying there LEDs can be paralleled, but they will not do it...and say that their leds should be in series.
On Osram website they specifically say that LEDs are manufactured primarily with light output in mind, and that because of this the tolerance on LED Vf is loose, and not to be relied upon.

I know that some places parallele LEDs, get a high field failure rate, but then just send a replacement lamp to the customer......they also use 1 Amp LEDs for <330mA operation , and use huge heatsinks.

Also, another point, in USA things are done properly.....in USA , 50% of all components in a product have to be made in USA, so you dont suffer savage cost-cuttting competition from the Far East
like we do here in Europe......so in USA you can afford to do conservative design, and have the product cost considerably more, and yet still sell it internally to USA customers.......in Europe, we cannot
do that.......yes i wish we could , but our Governments dont let us do that, shame , but true.


Also, in USA, you have LED foundrys in your own country, so you can be more sure that a "Cree" LED is a "Cree" LED.....IN Europe, a "Cree" LED could be a cheap fake from the Far East with lousy tolerance.

The day a Cree or Osram official person tells me that its absolutely fine to parallel any of there leds, then i'll think about doing it more than i do now, but they dont say such things.
I dont think they will any time soon, because the whole LED foundry fraternity is currently engaged in mostly trying to make LEDs give more light output per Watt.

Another point is that binned LEDs tend to be more expensive, as an example i would look at the Cree XPEHEW vs the rest of the XLamp range.

By the way, thanks for your above post Sunnyskyguy, i am still going through it in places and am thankful

- - - Updated - - -

Sunnyskyguy

- Steady state DC is ok but some designs shut down abruptly faster than the reverse 1MΩ typ leakage and junction xxx pF values.
- When pulsing LEDs the junction capacitance changes with voltage and off state and reverse voltage leakage resistance can vary significantly
- This RC parallel equivalent circuit forms a value in series with all the others. If they are all equal. The voltage declines equal. If unequal then the device with the highest resistance gets REVERSE BIASED.

..though for example the Cree XPEWHT has up to 5V reverse voltage withstand, and its hard to see this mechanism giving more than 5V of reverse voltage.

I must admit i'd never heard of this failure mode before...
 
Last edited:

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top