What types of driver circuits are there?

[balou]

The title pretty much sums it up... what different types of driver circuits are there, and how do they compare? And how valid are my observations to the little list I made below? Which types did I miss? Which types are used commercially, which are the most efficient?

Some I know of:
-Plain old resistor. Brightness changes with battery voltage, resistor heats up/wastes energy.
-PWM. Apparently not so good because LEDs have better efficiency at lower currents
-"BangBang", as described in a thread linked in this subforums sticky. PWM with filter to produce somewhat constant voltage and with feedback loop.
-Adjustable voltage power supply. In theory, just adjust voltage until the LED lets through your desired current. I guess there's some disadvantage... ?

 

[chimo]

This thread may answer most of your questions (and a few others you may not have asked).

The title pretty much sums it up... what different types of driver circuits are there, and how do they compare? And how valid are my observations to the little list I made below? Which types did I miss? Which types are used commercially, which are the most efficient?

Some I know of:
-Plain old resistor. Brightness changes with battery voltage, resistor heats up/wastes energy.
-PWM. Apparently not so good because LEDs have better efficiency at lower currents
-"BangBang", as described in a thread linked in this subforums sticky. PWM with filter to produce somewhat constant voltage and with feedback loop.
-Adjustable voltage power supply. In theory, just adjust voltage until the LED lets through your desired current. I guess there's some disadvantage... ?

 

[kramer5150]

Theres also this thread...
http://www.candlepowerforums.com/vb/showthread.php?t=188227

Oops sorry, looks like you are already aware of this one.

 

[SemiMan]

Really you need to talk about two things, current control and brightness control:

Current control:
- NONE: I.e. resistor, output varies with battery voltage. Generally inefficient.
- NONE: Filtered PWM . Generally efficient.
- Linear Regulator: LED voltage must be lower than battery voltage. LED current constant. Tends to be inefficient.
- Switching Regulator: LED voltage can be higher or lower than battery voltage. Efficiencies from 50% to 95%+ various implementations including: buck, boost, buck-boost, sepic, switch-cap, flyback, etc.

Brightness Control
- Change the resistor
- Filtered PWM: putting this in both. Really this is just a way of making a variable voltage. It is not closed loop.
- Current Variation: Can be used with Linear regulator or Switching regulator. Essentially you control what current goes to the LED.
- PWM: Can be used with resistor, linear or switching regulator. You vary the duty cycle to control the brightness.
- Note that Current variation and PWM can be used together

Semiman

 

[balou]

Ok, that answers why constant voltage is not a good idea. Vf varies by LED and by temperature. Thermal runaway.. :poof:

So this leads the thread in a new direction:
What types of constant current driver circuits are there?
Constant current sources is something you find almost nothing about... in textbook it's either a JFET-based few milliamp thingy, or LM317-based, with about 3.5v dropout voltage

Surely there are other topologies than what mpf described here:
http://www.candlepowerforums.com/vb/showthread.php?t=201383

edit: d'oh... waited few hours between reloading page. have to edit this post further, this was an answer to chimos first post..

 

[chimo]

Most of the constant current drivers you see here are derived by sensing a voltage drop over a small sense resistor in series with the LED. This voltage is fed back to the switcher chip in various manners.

Lots of useful information found here as well. Prepare to do some reading. :)

Ok, that answers why constant voltage is not a good idea. Vf varies by LED and by temperature. Thermal runaway.. :poof:

So this leads the thread in a new direction:
What types of constant current driver circuits are there?
Constant current sources is something you find almost nothing about... in textbook it's either a JFET-based few milliamp thingy, or LM317-based, with about 3.5v dropout voltage

Surely there are other topologies than what mpf described here:
http://www.candlepowerforums.com/vb/showthread.php?t=201383

edit: d'oh... waited few hours between reloading page. have to edit this post further, this was an answer to chimos first post..

 

[SkipH]

Most of the constant current drivers you see here are derived by sensing a voltage drop over a small sense resistor in series with the LED. This voltage is fed back to the switcher chip in various manners.

Lots of useful information found here as well. Prepare to do some reading. :)

:confused: The link above does not seem to work for me....can you access that web site?

Thanks

 

[chimo]

:confused: The link above does not seem to work for me....can you access that web site?

Thanks

I edited the post above to link to the non-mirrored site. It should be OK now.

 

[balou]

kramer5150: Unfortunately thats just a list of commercial offerings, nothing about the technology behind.

SemiMan: Well, if brightness is controlled by current, then there shouldn't be two different systems? But I think I get your point - I read about an AMC7135 board from DX, which uses a constant current IC which is PWM'ed by a µC.

chimo: Thanks for the info about the current sense resistor. Makes sense. And yes, I am prepared to do a lot of reading ;)
Unfortunately, link isn't working for me either. edit: works now


Soooo... it seems to me that a filtered PWM with current sense resistor is one of the most elegant and most efficient solutions (given that Vin > Vf). Or is it?


By the way: the reason I opened this thread is that I see people building and discussing the perfect flashlight, led, machining method etc,etc... but not the best way to drive a LED! That seems a little odd to me... imho the driver is a major component, influencing brightness and runtime. there is mostly discussion about premade circuits, and there are a few user-made circuits, but most of them are unfortunately black-box solutions - you don't get to know how it exactly works, nor is there source-code supplied.

In the past, I spent some time at head-fi... the custom amplifier section there is very interesting - people posting schematics and layouts of amplifiers, helping newbies with electronic design, and people improving one another's circuits. Unfortunately, I don't see anything like this here (or else please tell me where) - people seem to regard LED driver chips as trade secret. And unfortunately, economic pressure always seems to prevent the best solution - what is getting sold are 'acceptably working' solutions, with different levels of 'acceptable'. And judging by what non-CPFers find to be an acceptable amount of money for a flashlight... :(

 

[mpf]

it seems to me that a filtered PWM with current sense resistor is one of the most elegant and most efficient solutions (given that Vin > Vf). Or is it?

For Vin == (Vf + a little bit), in my judgement, it is for a small number of parallel stacks of leds.

The best way to drive leds depends of many things.
i) What is the power source, e.g. a single AA (1.2V - 1.5V), Lithium (3.3V), battery stack (3.3V - ??V)
ii) How many leds are being driven, e.g. 1, 4, 6, 28
iii) Is variable output needed. i.e. a single fixed current, a High and Low, or multiple levels. What is the lowest output level needed.
iv) How important is 'efficiency'. This is related to run time and heat dissipation of the controller components. Cost and size of the controller.


The latest design I am working on is a 60W 28die torch. The battery stack is 12 x AA NiMh. Using a design based on http://www.candlepowerforums.com/vb/...d.php?t=201383 would require 7 parallel stacks of 4 dies each. I became concerned about the current sharing between this many stacks of leds so I am looking at one stack of 28dies driven by a 100V boost converter or two stacks of 14 dies each driven by two separate 50V converters.

Currently the two separate 50V converters seems to be the way to go due to the simplier design and reduced dissipation per component.

The IC I am looking at is the LTC3783. This IC allows for both variable current control and PWM. At present I intend to use the variable current control to get a 10 to 1 current control, i.e. 60W to 6W and then use PWM to control the output from 6W down to 0.5W or less. Using current control down to 6W improves the efficiency of the Leds. Using PWM below this removes the problems of trying to measure the current at very low levels where circuit noise starts to be noticable.

The LTC3782's would be controlled by a Attiny uC which would provide the variable voltage setpoint to control the LTC3782's current and also provide the PWM for the very low level output control. The uC would also control the battery charging. For battery charging I will probably use a linear regulator as efficiency is not big issue and the expectied dissipation is small compared to the 60W the leds will put out.

matthew

p.s. You also need to consider what happens when you turn the torch off and put it in the cupboard for 12months. Will the batteries go flat due to self discharge or due the the residual current drawn by the uC.