Author Topic: LED driver circuit efficiency  (Read 10889 times)

0 Members and 1 Guest are viewing this topic.

Offline Federal Farmer

  • Contributor
  • Posts: 5
  • Country: us
    • GertWeb
LED driver circuit efficiency
« on: July 21, 2012, 05:00:04 PM »
Good day…

I scored a bunch of high power LED’s for free and have been looking at different drivers to power them with.
I picked up a kit at a local electronics chain store and reverse engineered it.
(If anyone is interested: http://www.gertweb.com/projects/images/LED%20Driver.pdf)



I built my own based on this design and took some measurements and found it's best efficiency was 75 to 78% @ 7.5 to 8v input.



Any suggestions on improving the above circuit or perhaps a better/different circuit altogether?
I was hoping to be at 80 to 90% efficient, or is this wishful thinking.
I have about 40 of these LED's so the cheaper the driver the better.

For an efficiency plot and other information visit my FB page.
http://www.facebook.com/media/set/?set=a.445878665445702.103623.100000707171935&type=3

EDIT 8/18/2012
Updated my personal web page to include this and other projects.
http://www.gertweb.com/arduino.html
« Last Edit: August 20, 2012, 09:19:54 AM by Federal Farmer »
 

Offline HackedFridgeMagnet

  • Super Contributor
  • ***
  • Posts: 1854
  • Country: au
Re: LED driver circuit efficiency
« Reply #1 on: July 21, 2012, 05:05:36 PM »
Hi Federal Farmer

There are people who don't use facebook, I am one of them.

But I am interested in Led Drivers, if you want to post your details somewhere else, maybe even in this thread.
 

Online IanB

  • Super Contributor
  • ***
  • Posts: 8152
  • Country: us
Re: LED driver circuit efficiency
« Reply #2 on: July 21, 2012, 05:16:01 PM »
It's going to be hard to get above 80% efficiency for most power converters without a lot of effort. That's about the sweet spot where most designs end up.
I'm not an EE--what am I doing here?
 

Offline Rufus

  • Super Contributor
  • ***
  • Posts: 2095
Re: LED driver circuit efficiency
« Reply #3 on: July 21, 2012, 10:32:03 PM »
It's going to be hard to get above 80% efficiency for most power converters without a lot of effort. That's about the sweet spot where most designs end up.

Efficiencies in the mid 90s are not difficult for LED buck switchers. Harder if you are restricted to building on proto board.

If the OP is using exactly the circuit posted it will be loosing 10% or more just in the input bridge rectifier.

 

Offline Federal Farmer

  • Contributor
  • Posts: 5
  • Country: us
    • GertWeb
Re: LED driver circuit efficiency
« Reply #4 on: July 22, 2012, 01:13:12 AM »
Hi Federal Farmer

There are people who don't use facebook, I am one of them.

But I am interested in Led Drivers, if you want to post your details somewhere else, maybe even in this thread.
Off Subject…
Facebook or something like it will become the government of the future.
Check out http://fora.tv/2008/08/08/Daniel_Suarez_Daemon_Bot-Mediated_Reality




Measuring the performance characteristics of the constant current LED driver.
 
At a “good” brightness level I measured 7.85 volts on the input and 6.12 volts on the output.
Also measuring the current at this level I recorded 92mA. This gives an efficiency of 77.96%
 (100 * 6.12) / 7.85 = 77.96%
 
At an “excellent” brightness level I measured 8.71 volts on the input and 6.57 volts on the output.
Also measuring the current at this level I recorded 284mA. This gives an efficiency of 75.43%
 (100 * 6.57) / 8.71 = 75.43%
 
The graph below is a plot of measured values showing input and output voltages verses current.
At the highest measured input voltage of 11 volts the circuit is 66% efficient.
Like all things in the physical world there are diminishing returns for applied effort.
There are religious aspects to science it is just a matter of observation and awareness. (I digress)

If anyone needs or wants help with graphing in Excel let me know.





This is actually my first proto build.
I used mostly the trimmed leads for the traces.
In hindsight I should have just bent them over and soldered them in place. (live and learn)

Q: Is it normal to mount the comparator in a socket?
I am thinking the original was designed this way because it has a shorter lifespan and this makes it easily replaceable.
Or is it just to avoid putting heat on a sensitive component? (thoughts or comments on this are welcome)






« Last Edit: July 22, 2012, 01:55:12 AM by Federal Farmer »
 

Offline Federal Farmer

  • Contributor
  • Posts: 5
  • Country: us
    • GertWeb
Re: LED driver circuit efficiency
« Reply #5 on: July 22, 2012, 01:46:22 AM »
If the OP is using exactly the circuit posted it will be loosing 10% or more just in the input bridge rectifier.

Removing the diodes was one of my first thoughts too.
The original board was designed for either AC or DC input.

What would bumping up the resistors to a higher wattage rating do?
My thoughts are there would be less heat generated giving better efficiency.

For what it is worth...
I can buy the components for a stripped down version of this circuit for about $10.00 from digi-key.
or I can buy completed assemblies of a similar circuit for $5.00 (http://www.parts-express.com/pe/showdetl.cfm?Partnumber=073-050)

But then what would I have learned?  ;)

Thanks to all for sharing your wisdom and thoughts.
 

Offline EEMarc

  • Regular Contributor
  • *
  • Posts: 94
  • Country: us
Re: LED driver circuit efficiency
« Reply #6 on: July 22, 2012, 02:17:04 AM »
http://xkcd.com/833/

The socket is fine. It isn't a device that requires power dissipation through the pins or has any substantial current running through it.

You're loosing most of your efficiency through the 4 diodes making the bridge as Rufus mentioned. If you want to improve the efficiency with a DC source, hook power up directly to V+ and ground. If you want to keep the diodes in and improve their efficiency, you can change them to low forward voltage drop Schottky diodes.

The other efficiency gains are smaller such as component selection tradeoff for higher efficiency at the cost of higher price and larger sized components. Low cost was used for the inductor which is probably contributing a fair amount to the loss in efficiency. Also, the current sense circuit can be made to lose less by reducing it's resistance. This will require a lower fixed voltage to the comparator to compensate for the lower voltages and will reduce stability margins.
 

Offline Federal Farmer

  • Contributor
  • Posts: 5
  • Country: us
    • GertWeb
Re: LED driver circuit efficiency
« Reply #7 on: July 22, 2012, 02:45:36 AM »
http://xkcd.com/833/ - LOL ::)

If you didn’t know you are looking at voltage and current, labels probably would not help you.
But point taken.

"Or is it just to avoid putting heat on a sensitive component?"
I was thinking in terms of the soldering process.
« Last Edit: July 23, 2012, 12:32:17 AM by Federal Farmer »
 

Offline madires

  • Super Contributor
  • ***
  • Posts: 3711
  • Country: de
  • A qualified hobbyist ;)
Re: LED driver circuit efficiency
« Reply #8 on: July 22, 2012, 03:03:38 AM »
For what it is worth...
I can buy the components for a stripped down version of this circuit for about $10.00 from digi-key.
or I can buy completed assemblies of a similar circuit for $5.00 (http://www.parts-express.com/pe/showdetl.cfm?Partnumber=073-050)

There are a lot a buck converters with a current limiter for driving LEDs. If you like to stay with DIP have a look at LM3402/LM3404/LM3405.
FidoNet 2:240/1661
 

Offline EEMarc

  • Regular Contributor
  • *
  • Posts: 94
  • Country: us
Re: LED driver circuit efficiency
« Reply #9 on: July 22, 2012, 03:31:23 AM »
I've never had a problem with overheating the pins on an IC. Then again, I don't use unleaded solder that takes longer to wet out. Sockets are usually just the easy way to replace bad ICs. Usually they don't go bad, but sometimes bad stuff happens such as a voltage spike or piece of metal finds a way to short stuff out.

As for graphs labels, there is something called the curse of knowledge. When someone is immersed into a project, they have prior knowledge and require less and less information to know exactly what is going on. They save time by putting less work into labeling and other such tasks. This can be problematic when someone new comes in to help and tries to figure out what is going on because they don't the prior knowledge. The curse of knowledge is having prior knowledge and assuming that other have the same information when they don't. Plus it is helpful to have rock solid information to go on as opposed to assumptions.

More importantly, whenever I get the setup that well for an XKCD insertion, I take it.
 

Offline HackedFridgeMagnet

  • Super Contributor
  • ***
  • Posts: 1854
  • Country: au
Re: LED driver circuit efficiency
« Reply #10 on: July 22, 2012, 10:55:59 AM »
Hi Federal Farmer

There are a few other small things worth pointing out.
R4 doesn't need to be so large, you probably don't even need it, it is only for dissipating the energy from the gate in each switch. With that large package fet, I don't think you need it.

It is also important to match Q1, Q2 and Q3 so that the bjts change charge up/down the gate as fast as possible.
I am not an expert on this, but I have been looking into it for a while.

After choosing your topology, I think the general plan is select your fet next. You want a safe voltage margin, (100v) is probably excessive, low RDS on and a small gate charge. I ignore the power or current rating as I am trying to get the absolute minimum power loss anyway, so it should be running fairly cool anyway.
Obviously N Channel fets can give better rds(on) values here, but you would have to flip the topology about.

You are using what is often called a Hysteretic buck converter, the hysteris should be in the comparator, it's primary advantage is that it minimises the number of switches for the amount of ripple. Therefore can be very efficient.
It's chief disadvantages are you can't easily control the minimum switching rate, so you may get flicker, especially if you add a dimmer.

The BJTs might need a bit more oomph (Imax) to drive a large fet, (large gate charge), maybe BC639, BC640 they were suggested to me.
Another thing to watch is you don't want to drive either of the BJTs into heavy saturation, so you should limit the base current so it just goes into saturation, and then it can come out of saturation faster.
You can model these things in LTSpice and look at the waveforms on a cro, what you want are nice square edges, as much as possible.

Two other lossy places are the current sense resistors, can they somehow be smaller?
and the inductor.
You might be able to use a smaller inductor depending on the allowable ripple, the smaller inductor will have lower series resistance generally.
The bigger the footprint too will give lower series resistance.
You can keep reducing until you are hitting the absolute maximum led current. While in development I would put fuses on the leds as they are easy to blow.

Probably the most efficient thing you can do is somehow use a lot of Leds in series, this would require a higher voltage circuit and a redesign, with a lot of leds in series you can easily get 90%.

Here is a link to my led project
http://www.eevblog.com/forum/projects-designs-and-technical-stuff/domestic-led-lighting-project/

Good luck.

ps What sort of leds?
 

Online IanB

  • Super Contributor
  • ***
  • Posts: 8152
  • Country: us
Re: LED driver circuit efficiency
« Reply #11 on: July 22, 2012, 11:04:40 AM »
You might be able to use a smaller inductor depending on the allowable ripple, the smaller inductor will have lower series resistance generally.

For a given value inductance you can reduce the resistance by using thicker wires. It may require a bigger core opening to accommodate the extra wire thickness. If you are building a one-off rather than a production design you can wind your own inductor on a standard core and make the resistance as low as you want it to be.

Quote
Probably the most efficient thing you can do is somehow use a lot of Leds in series, this would require a higher voltage circuit and a redesign, with a lot of leds in series you can easily get 90%.

I don't get this. The efficiency of boost converters generally reduces as the voltage ratio becomes greater. Putting many LEDs in series would require a greater output voltage. I can only see this lowering the efficiency.
I'm not an EE--what am I doing here?
 

Offline Rufus

  • Super Contributor
  • ***
  • Posts: 2095
Re: LED driver circuit efficiency
« Reply #12 on: July 22, 2012, 12:04:15 PM »
The smaller the inductor (inductance not physical) the higher the switching frequency and so the higher the switching losses.

I * R and I * V losses in the MOSFET, inductor, sense resistor, and free wheeling diode are shared between all LEDs in a chain so longer chains are more efficient as long as you stay buck. Going boost means peak currents get much higher requiring bigger components and losses from higher rms currents.

There really isn't any point trying to make a good LED switcher with discretes. There are too many good switcher chip solutions. I like the Allegro A6210. Minimal external components with an efficient topology, up to 46v supply and an integrated 3A MOSFET switch for not much more than the cost of the MOSFET in the OP's circuit. Like a lot of good chips only available in QFN so you really need a PCB.
 

Offline digsys

  • Supporter
  • ****
  • Posts: 1558
  • Country: au
    • DIGSYS
Re: LED driver circuit efficiency
« Reply #13 on: July 22, 2012, 12:16:01 PM »
Quote
I don't get this. The efficiency of boost converters generally reduces as the voltage ratio becomes greater. Putting many LEDs in series would require a greater output voltage. I can only see this lowering the efficiency.
You can "tune" or "optimise" ANY voltage ratio to be at the highest efficiency for any power ratio. Usually some form of bell curve.
The good thing about series LEDs is the Vout and Iout are pretty much fixed. PLUS you save a lot on individual losses.
Hello <tap> <tap> .. is this thing on?
 

Online IanB

  • Super Contributor
  • ***
  • Posts: 8152
  • Country: us
Re: LED driver circuit efficiency
« Reply #14 on: July 22, 2012, 01:02:09 PM »
You can "tune" or "optimise" ANY voltage ratio to be at the highest efficiency for any power ratio. Usually some form of bell curve.
The good thing about series LEDs is the Vout and Iout are pretty much fixed. PLUS you save a lot on individual losses.

But the power ratio is the efficiency. If you are aiming for the highest absolute efficiency at a given power output the voltage ratio will limit what you can achieve. For example if you want to boost 2 V to 4 V at 700 mA output you will be able to design a higher efficiency than boosting 1 V to 4 V at the same 700 mA output.

With a buck configuration you will likewise be able to design the best efficiency if the voltage ratio is closer to 1:1. That's why people above suggested putting more LEDs in series. You win until your output voltage starts needing to be greater than your input voltage.
I'm not an EE--what am I doing here?
 

Offline digsys

  • Supporter
  • ****
  • Posts: 1558
  • Country: au
    • DIGSYS
Re: LED driver circuit efficiency
« Reply #15 on: July 22, 2012, 01:24:51 PM »
Quote
But the power ratio is the efficiency. If you are aiming for the highest absolute efficiency at a given power output the voltage ratio will limit what you can achieve. For example if you want to boost 2 V to 4 V at 700 mA output you will be able to design a higher efficiency than boosting 1 V to 4 V at the same 700 mA output.
With a buck configuration you will likewise be able to design the best efficiency if the voltage ratio is closer to 1:1. That's why people above suggested putting more LEDs in series. You win until your output voltage starts needing to be greater than your input voltage.
Having worked with Solar EV teams for many years, we don't design ANYTHING under 98.5% efficiency, and have no problems doing so.
That includes Motor controller, MPPTs, power converters. We have ratios of 4:1 -> 1:2. Often had TOP of the bell curve at 99.0 - 99.2%
I admit, our buck/boost converters are a lousy ~96-97% just past the peak .. been tinkering with multi-topography, which are becoming
popular. There are some new "complete" devices which have the switching FETs AND synchronous diodes built in with high efficiencies.
Power conversion is improving in leaps and bounds.
Hello <tap> <tap> .. is this thing on?
 

Offline EEMarc

  • Regular Contributor
  • *
  • Posts: 94
  • Country: us
Re: LED driver circuit efficiency
« Reply #16 on: July 22, 2012, 04:38:40 PM »
You're giving me some Eta envy there digsys. I've done a fair amount of converter work but I usually design efficiencies for low to mid 90's.
 

Offline SeanB

  • Super Contributor
  • ***
  • Posts: 14482
  • Country: za
Re: LED driver circuit efficiency
« Reply #17 on: July 22, 2012, 06:28:19 PM »
Digisys, can you post some of your designs that are not encumbered by copyright issues? This would be a good thing to see, and would probably improve all the designers here a lot.
 

Offline G7PSK

  • Super Contributor
  • ***
  • Posts: 3490
  • Country: gb
  • It is hot until proved not.
Re: LED driver circuit efficiency
« Reply #18 on: July 22, 2012, 06:41:22 PM »
Q: Is it normal to mount the comparator in a socket?
I am thinking the original was designed this way because it has a shorter lifespan and this makes it easily replaceable.
Or is it just to avoid putting heat on a sensitive component? (thoughts or comments on this are welcome)

The socket is most likely used in the original kit so that ham fisted kids with a 100 watt soldering iron that was inherited from granddad don't cook the IC, used to see a lot of kits like that in the eighties and nineties.
 

Online IanB

  • Super Contributor
  • ***
  • Posts: 8152
  • Country: us
Re: LED driver circuit efficiency
« Reply #19 on: July 22, 2012, 06:44:57 PM »
Yeah, I think it's "traditional" to use IC sockets in kits. One reason being that if you mess it up it (for example putting it in the wrong way round) it is much harder to unsolder and remove a DIP package than a small two lead component.
I'm not an EE--what am I doing here?
 

Offline digsys

  • Supporter
  • ****
  • Posts: 1558
  • Country: au
    • DIGSYS
Re: LED driver circuit efficiency
« Reply #20 on: July 22, 2012, 08:00:48 PM »
You're giving me some Eta envy there digsys. I've done a fair amount of converter work but I usually design efficiencies for low to mid 90's.
Digisys, can you post some of your designs that are not encumbered by copyright issues? This would be a good thing to see, and would probably improve all the designers here a lot.
These folks were original members of our club many years ago and pioneers of development work. They were regarded as the best on the
planet, and Tritium is still the best there is. All designed in Queensland.
www.aerl.com.au   Stuart passed away a couple years ago sadly (from LEAD poisoning in solder !). We have a new team designing MPPTs, but I
can't say anything about that yet, as we're gearing up for a new solar challenge OCT next year, all new classes.
https://www.facebook.com/events/396315923739572/  and  www.worldsolarchallenge.org
www.tritium.com.au/products/TRI88/index.html   Used in pretty much most Solar race cars and serious competition EV cars.
There are 1-2 open source projects going on around the planet on power converters, ~98% efficiency but with a flatter bell curve.
I'll dig up some links, and past stuff when I find a minute. Also need to check what we have public domain.
If you google "Solar" "race" "MPPT" etc you should find the public forums, and I can't get into trouble :-)
As for the other power converters, I'd need to check what are "race secrets". Hopefully that's enough to start chewing on.
Hello <tap> <tap> .. is this thing on?
 

Offline HackedFridgeMagnet

  • Super Contributor
  • ***
  • Posts: 1854
  • Country: au
Re: LED driver circuit efficiency
« Reply #21 on: July 22, 2012, 08:05:11 PM »
Quote
Quote
    You might be able to use a smaller inductor depending on the allowable ripple, the smaller inductor will have lower series resistance generally.

For a given value inductance you can reduce the resistance by using thicker wires. It may require a bigger core opening to accommodate the extra wire thickness. If you are building a one-off rather than a production design you can wind your own inductor on a standard core and make the resistance as low as you want it to be.
Sorry for the ambiguity, as you already know, I meant smaller inductance for lower series resistance and in other places I meant larger physical size for lower series resistance.


Quote
Quote
    Probably the most efficient thing you can do is somehow use a lot of Leds in series, this would require a higher voltage circuit and a redesign, with a lot of leds in series you can easily get 90%.
I don't get this. The efficiency of boost converters generally reduces as the voltage ratio becomes greater. Putting many LEDs in series would require a greater output voltage. I can only see this lowering the efficiency.
Here I meant increasing the rail voltage at which the fet switches and still use a buck converter. This would mean a complete redesign.
The saving I was suggesting here, was as I think Digisys mentioned, you save a lot by sharing the the series losses and control losses between the all the leds.

Those efficiency figures Digisys sound out of this world, are they high power?
Your post just beat mine, I think I have heard of this company and this man. Sorry he passed away.

 

Offline digsys

  • Supporter
  • ****
  • Posts: 1558
  • Country: au
    • DIGSYS
Re: LED driver circuit efficiency
« Reply #22 on: July 22, 2012, 08:17:15 PM »
Quote
Those efficiency figures Digisys sound out of this world, are they high power?
Your post just beat mine, I think I have heard of this company and this man. Sorry he passed away.
Most designs are between 300-600W a module, designed around a section of solar array. The technology itself can
naturally be used in any type of converter. I forgot to add, they work on Inputs of ~24-100V+ AND Outputs of 80-180V
with ANY combination.
As for the Tritium motor controller go up to 60KW+ (There's one on Swinburne's Formula SAE Electric race car.)
teamswinburne.com  which we're also active with.
Hello <tap> <tap> .. is this thing on?
 

Offline HackedFridgeMagnet

  • Super Contributor
  • ***
  • Posts: 1854
  • Country: au
Re: LED driver circuit efficiency
« Reply #23 on: July 22, 2012, 08:59:43 PM »
Yes it is the same product, that I did see before.

I remember the product name now, CoolMax.
The quality of the PDFs has certainly improved since I last looked.
Last time they had all this saturated Blue and Red text which hurt my eyes.
I am definitely interested in what you do though, I will read up.




 

Offline Psi

  • Super Contributor
  • ***
  • Posts: 6278
  • Country: nz
Re: LED driver circuit efficiency
« Reply #24 on: July 22, 2012, 10:07:12 PM »
Those efficiency figures Digisys sound out of this world, are they high power?

Most of the general purpose switch-mode supplies are made to the lowest price (efficiency is expensive) but there are plenty of high efficiency ones available.

It's not uncommon to have >95% efficiency switch-mode, you just have to design it that way and use really good spec'ed parts. eg, ultra low rsd(on) fets, synchronous rectification, 400+kHz inductors.

Here's a carpc powersupply i bought a while back.
It's >95% efficient and does 6-24VDC input to 12V @ 12A output (both step up and step down)
http://store.mp3car.com/DSX12VD_140W_DC_DC_Regulator_p/pwr-025.htm

The inductor design is quite interesting, it's flat copper.

« Last Edit: July 22, 2012, 10:19:50 PM by Psi »
Greek letter 'Psi' (not Pounds per Square Inch)
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf