Also, one big problem you will encounter is this:
If you lowpass filter the PWM signal to get an analog control voltage, and then apply that to the LEDs, you will get illumination that is exponentially related to the voltage. Your illumination vs voltage will look like the current vs voltage curve of the LED. Is that what you want? I doubt it.
If you use a voltage to current converter, you will get linear response, but you will never reach the full 3.3V across the LEDs.

 

Use current control to drive an LED string, never use voltage control. There are many great constant current LED drivers out there, such as Diodes Inc., Natl Semi, TI, Supertex, etc.

The reason nobody drives an LED with voltage is because that is an awful way to do it. It is thermally unstable, and the resolution is poor at best. Current drive results in thermal stability and good resolution over a broad range.

I have nothing against examining new ways of doing things, but LEDs have been around a long time, and the voltage drive method has proven to be an utter disaster. Nothing new at all about voltage drive. It works great for incandescent lamps, but lousy for LEDs.

Use current drive for LEDs.
Use current drive.
LEDs are current-driven devices, they are not voltage driven. It's in the nature of their physics. A constant voltage source directly across an LED results in a current given by:

If = Is(T)*exp[(Vf/Vt)-1], where Is(T) increases greatly with temp "T". Forcing a voltage across the LED results in power dissipation, resulting in higher T, resulting in higher Is, resulting in higher If, resulting in higher power, resulting in higher T, resulting in higher Is, ad infinitum.

When current driven:

Vf = Vt*ln[(If/Is(T))+1].

A current in the LED results in a voltage per above equation, resulting in power, temp rise, then an increase in Is. But Is is in the denominator so that Vf is decreased. Thermal stability is reached since the temp rise increment converges.

When driven with voltage, the LEDs are thermally unstable. When driven with current, they are stable. No need to discuss drive methods any further.

For LEDs current drive is the only viable method. Under no circumstances should an LED string be directly driven from a voltage source.

Claude

 

Cabraham, if you read the entire thread, you will see that you are whipping a dead horse. You have added 1 to the number of people that our OP is disgusted with.

 

Cabraham, if you read the entire thread, you will see that you are whipping a dead horse. You have added 1 to the number of people that our OP is disgusted with.

Yes the current drive thing was mentioned earlier, but I thought it wouldn't hurt to quantify it with equations. I hope the OP isn't angry because I was only explaining the why as to using current drive.

Rather than tell somebody "you must do it this way", I prefer to give the reason. I did not intend to offend anyone. Hopefully we can move past this.

Claude

 

Yes the current drive thing was mentioned earlier, but I thought it wouldn't hurt to quantify it with equations. I hope the OP isn't angry because I was only explaining the why as to using current drive.

Rather than tell somebody "you must do it this way", I prefer to give the reason. I did not intend to offend anyone. Hopefully we can move past this.

Claude

I agree with you, Claude. I think our OP understands the physics, at least to some degree. I think his 3.3V power supply and limited board space are the root causes of his frustration. He was looking to us for help, and none was forthcoming that would work for him.

 

Also, one big problem you will encounter is this:
If you lowpass filter the PWM signal to get an analog control voltage, and then apply that to the LEDs, you will get illumination that is exponentially related to the voltage. Your illumination vs voltage will look like the current vs voltage curve of the LED. Is that what you want? I doubt it.

Now this is the kind of input I like - as it goes, our eyes aren't linear so it may well be the LED response I need/want.

If you use a voltage to current converter, you will get linear response, but you will never reach the full 3.3V across the LEDs.

This is not a problem - in fact that's my point - if you supply less voltage to a LEDS vs their intended rated fwd voltage you still get good brightness but without straying into the danger zone of thermal runaway & the likes.

Even if the fade response using PWM is not what I seek, I could bin PWM & instead use a digital pot (under MCU control) as the input to the voltage folower. ...I'm going to order up a couple of AD8591s (as it looks like a possible contender) & have a dabble.

A useful maxim - new on here doesn't always = n00b without a clue ...so please, less of the patronising, smug input. I'm well aware that an LED is best viewed as a current device (vs. voltage), but for my end applictaion PWM'ing a fixed voltage through a resistor neither gives the LED fade characteristic I seek,....but worse still as the LED at the tail end of its fade jitters between on & off, this causes faint audible clicking - I don't have sufficient PCB real estate to put in place workarounds (I would if this was the only issue, but like I say the fade response is not ideal anyway, so exploring creative alternatives, which while breaking convention may ultimately give the end result I need)

 

Quote:
Originally Posted by Ron H
If you use a voltage to current converter, you will get linear response, but you will never reach the full 3.3V across the LEDs.

You replied:

This is not a problem - in fact that's my point - if you supply less voltage to a LEDS vs their intended rated fwd voltage you still get good brightness but without straying into the danger zone of thermal runaway & the likes.

In post #4, you said this:

I was hoping to get a solution here that would follow the input voltage 0->3.3V but with a VCC of 3.3V.

I spent a lot of time trying to design a follower that would go within less than 100mV of the rail with a 150mA load (not sure where I got this number). I didn't notice the "<" in you voltage statement. Just how much is "<"? Post #4 does tell us that you think 0.7V is too much.
I don't think a follower will solve your problem, but just for grins, why don't you tell us the maximum simultaneous voltage and current requirements you think you need. A V-I converter might be able to solve your problem, if it can have enough voltage headroom.

 

You replied:

In post #4, you said this:
I spent a lot of time trying to design a follower that would go within less than 100mV of the rail with a 150mA load (not sure where I got this number). I didn't notice the "<" in you voltage statement. Just how much is "<"? Post #4 does tell us that you think 0.7V is too much.
I don't think a follower will solve your problem, but just for grins, why don't you tell us the maximum simultaneous voltage and current requirements you think you need. A V-I converter might be able to solve your problem, if it can have enough voltage headroom.

I ultimately want this to be a battery solution. I am using 3.4V LEDs but running them at 3.3V or less ...I can allow the voltage rail of the driving device to be a little higher (therefore ensuring the output gets to 3.3V).

The maximum load will be 240mA.

 

I ultimately want this to be a battery solution. I am using 3.4V LEDs but running them at 3.3V or less ...I can allow the voltage rail of the driving device to be a little higher (therefore ensuring the output gets to 3.3V).

The maximum load will be 240mA.

So (3.4V-3.3V)=100mV is the required headroom?

 

So (3.4V-3.3V)=100mV is the required headroom?

I guess what I'm saying is that I have a degree of flexibility with respect to the VCC of the driving device(so up to 300mV headroom)....my only big constraint is that I want the LEDs to run at about 3.3V or less.

 

I guess what I'm saying is that I have a degree of flexibility with respect to the VCC of the driving device(so up to 300mV headroom)....my only big constraint is that I want the LEDs to run at about 3.3V or less.

I have a current mode driver design almost ready for you. It would be controlled by the lowpass filtered PWM signal. This will avoid the unwanted brightness at low duty cycles which I believe is caused by persistence of vision or some related nonlinear response of vision.
If you are interested, I'll try to finish it. It's actually a common technique. Due to your low supply voltage, the devil is in the details.

 

whilst i'm not saying i will ultimately use it, it would be nice to see what this entails (not just for me but others on this forum), so if it's nearly done, then yes ...please post if you finish it.

 

whilst i'm not saying i will ultimately use it, it would be nice to see what this entails (not just for me but others on this forum), so if it's nearly done, then yes ...please post if you finish it.

OK, here it is. As I said, the topology is not new, but the devil is in the details.
In order for the current to be zero when the input is zero, the op amp's input offset voltage has to be very low. The output should also be able to swing rail-to-rail in order to be certain that the MOSFET can be driven over the full range of input voltage. The input range has to include the negative rail.
The MOSFET has to have a very low threshold voltage, and low Rds(on), in order to Be able to conduct 240mA with Vgs<3.3V And Vds<≈200mV.
That is why I picked these particular parts. You can choose others, so long as they meet these requirements. Other combinations of op amp and MOSFET might have stability problems, requiring a couple of added resistors and a capacitor for stability. I think (but can't guarantee) that the circuit as shown will not oscillate.
The output should go to 240mA when the input is full scale.

 

Thanks Ron.

My main worry with that circuit is the potential for an audible click as the mosfet biases off (because that 'switchover' will happen when the audio signal is lowest & therefore such low level audio won't 'mask' any potential audible bias off related click) ...There'll potentially be quite a few LEDs in play here (& therefore pulling a bit of current even at low brightness levels)...therefore I'd rather have a bit of current flowing even if the LEDs aren't visibly on. vs no current and the risk of a click (my PIC can attend to ensuring the current draw is zero after a prolonged absence of audio).

I did an experiment with a pink 3mm high brightness LEDs

Full LED brightness is about 3.4V ...and no perceivable visual brightness around 2.3V ...but significantly it had eactly the fade characteristic I was after ...in other words a graceful fadeout (vs PWM'm a fixed voltage through a resistor at a PWM of 1/1024th, which although dim is just too bright still)...so I do believe a voltage follower is the way to progress this.

I had imagined that I'd likely end up with a circuit where the ouput of the opamp would go to the LED anodes & the common cathodes grounded, with the incoming filtered PWM varying between 2.3V & 3.3V (which if I have my PIC VCC @3.3V, there should still should be decent resolution with 10 bit PWM)

 

Thanks Ron.

My main worry with that circuit is the potential for an audible click as the mosfet biases off (because that 'switchover' will happen when the audio signal is lowest & therefore such low level audio won't 'mask' any potential audible bias off related click) ...There'll potentially be quite a few LEDs in play here (& therefore pulling a bit of current even at low brightness levels)...therefore I'd rather have a bit of current flowing even if the LEDs aren't visibly on. vs no current and the risk of a click (my PIC can attend to ensuring the current draw is zero after a prolonged absence of audio).

I did an experiment with a pink 3mm high brightness LEDs

Full LED brightness is about 3.4V ...and no perceivable visual brightness around 2.3V ...but significantly it had eactly the fade characteristic I was after ...in other words a graceful fadeout (vs PWM'm a fixed voltage through a resistor at a PWM of 1/1024th, which although dim is just too bright still)...so I do believe a voltage follower is the way to progress this.

I had imagined that I'd likely end up with a circuit where the ouput of the opamp would go to the LED anodes & the common cathodes grounded, with the incoming filtered PWM varying between 2.3V & 3.3V (which if I have my PIC VCC @3.3V, there should still should be decent resolution with 10 bit PWM)

You have misinterpreted the circuit. The MOSFET doesn't switch. See the big lowpass filter on the input? That creates a very clean analog signal which goes from zero to 48mV on the op amp's + input (call this Vpos) as the PWM duty cycle goes from 0 to 100%. Due to feedback, this exact voltage appears on the - input, which means it is impressed across the 0.2Ω resistor. Therefore, the MOSFET source and drain (and LED) current is Vpos/0.2Ω, with NO switching. The current is totally analog.

 

Here is a voltage driver that should do what you want to try. The voltage across the LED should go from 2.3V to 3.3V as the duty cycle changes from 0 to 100%. Note that it uses the full PWM range. If you want to use only the upper part of the range, eliminate R1.
If it burns out your LEDs, I am not responsible.
Note that if one burns out, you will get an avalanche effect as the current in the remaining ones increases. Of course, this can happen with parallel LEDs with current mode drive too.
I didn't try to optimize the filter for response time vs ripple. If you like the way this works, I can try to do that if you want.

 

You have misinterpreted the circuit. . Due to feedback, this exact voltage appears on the - input, which means it is impressed across the 0.2Ω resistor. Therefore, the MOSFET source and drain (and LED) current is Vpos/0.2Ω, with NO switching. The current is totally analog.

Indeed I have.

The 48mV you mention which I can see is derived from the potential divider R3 & R6 (then smoothed)

If that's presented at the LED cathode, then this means the voltage across the LED is only going to vary between 3.3V & 3.252V???

I was seeking a circuit that varied between 2.3V & 3.3V at its output.....

 

Indeed I have.

The 48mV you mention which I can see is derived from the potential divider R3 & R6 (then smoothed)

If that's presented at the LED cathode, then this means the voltage across the LED is only going to vary between 3.3V & 3.252V???

I was seeking a circuit that varied between 2.3V & 3.3V at its output.....

Read the post again. The LED current varies from 0 to 240mA as the PWM duty cycle varies from 0 to 100%. This is because the voltage across the 0.2Ω resistor varies from 0 to 48mV (48mV/0.2Ω=240mA). That current passes through the MOSFET and the LED.
The LED voltage is not controlled. It follows the VI curve of the LED.

Also, see post #56 (you might have missed it).

 

Yes, you're right, I did miss your post #56!

Here is a voltage driver that should do what you want to try. The voltage across the LED should go from 2.3V to 3.3V as the duty cycle changes from 0 to 100%. Note that it uses the full PWM range. If you want to use only the upper part of the range, eliminate R1.
If it burns out your LEDs, I am not responsible.
Note that if one burns out, you will get an avalanche effect as the current in the remaining ones increases. Of course, this can happen with parallel LEDs with current mode drive too.
I didn't try to optimize the filter for response time vs ripple. If you like the way this works, I can try to do that if you want.

Great stuff (a hearty thanks - seriously)...this is the circuit I ultimately want to try (I understand your first circuit now...very clever, but I think it heavily depends on knowing how many LEDS are going to deployed......I was seeking a circuit that would allow a degree of flexibility without changing out components to suit)

I'm not worried about trashing a few LEDS (and at the risk of rustling feathers again...I don't see this are much risk, since I'll be running a heap of LEDs that are spec'ed for 3.4V at 3.3V or less...this is more a journey of interest! I'm particularly impressed with bringing R1 into play to form a potential divider with the PWM output (thereby allowing full PWM resolution - excellent stuff!)

The first part of the circuit I understand, but could you talk me through the opaamp mosfet part, because I'm just not locking on to the inner workings there!

 

Keep us posted on your progress (or lack thereof).