I am putting the blame on the consumer that will try to save and clearly don't care about quality.

I'd put the blame on companies that design crappy products. You can't expect a consumer to understand enough about how things to work to care; or be able to tear something down and know poor design from good design. Consumers should expect that whatever they buy will perform as the product implied.

 

How do you figure that PWM is constant current?
it varies the average voltage, not the current.

I use PWM made for LEDs. Here is an example. The Feed Back voltage is 0.1 volt. Current is measured by 0.1V across RS. The PWM does not know what the voltage is across the LED.

Here is another way. The IC knows the current in L1 by measuring the voltage across R1. The current in L1 is related to the current in the LEDs. Not great regulation but it works.

 

If you had a single 2V LED what voltage overhead would you consider to be reasonable with a series resistor?

If you had a hundred 2V LEDs in series what voltage overhead would you consider to be reasonable with a single series resistor?

That depends.

What is the acceptable range of currents in the LED? If it only has to be within 50%, the answer is different than if it has to be within 10%.

What is the typical and maximum Vf of the LEDs (and, if available, the minimum)?

 

To appease @WBahn how about I revise the required overhead to Vf x √n.

Again, the issue isn't Vf, but the uncertainty in Vf.

Imagine that tomorrow someone put on the market a brand new type of LED that had wonderful optical properties, but it's nominal Vf was 100 V and the min and max varied from that by at most 1 V. Would you still say that to get decent current control you need 100 V of overhead for a single LED?

 

I use PWM made for LEDs. Here is an example.

Okay,
You are using PWM in a switching converter configured with feedback to generate a constant-current output.
That makes sense.

Most applications of PWM are used to vary voltage.

 

Most applications of PWM are used to vary voltage.

True.

Battery chargers are CC with over voltage limiting. (many have temperature regulation.) We often regulate at 100A, when the battery is cold and as the temperature climes, we regulate at 20C rise. Temperature is in the loop.

MPP regulators adjust the duty cycle for max power out of the solar cell. The output current and voltage is not in the main loop.

We have heater/cooers that PWM to keep the temperature right. Voltage & current are not regulated.

Motor drivers (PWM) have speed or position or torque in the loop. Servos may use PWM to hold a position. When the motor is far from the right place the duty cycle is at 100% then as the position gets close the duty cycle reduces ending in 0%.

"PWM" describes how the power switch is being used. (linear vs digital) There are many subcategories, constant frequency vs variable, constant on time or constant off time, LLC which is variable frequency which has several different types. etc.

Flashlights (torches), I have made some where the current is 200% normal, then we watch the LED temperature. So 80C is what we are regulating on. This way we can say we are 2x brighter than other lights. And I know we are not over driving the LED because we watch it very carefully. Most likely our LEDs will last better than other lights.

 

@ anyone who cares:
Vs (Supply Voltage) should be a few volts MINIMUM above Vf. I said "MINIMUM". That is to say you need "Enough" voltage to light the LED's AND employ a resistor to limit the current in some reasonable fashion. There ARE cases where using a resistor is less desirable than a CC source, but for the sake of argument about Vs - you need enough voltage to push the current through the LED's.

Now; what about Maximum voltage? IN THEORY there is no limit. You can "IN THEORY" power a single 2Vf LED with 100kVDC. You'll just need a hell of a resistor or a hell of a constant current source. You don't "Always" double the Vf. You merely exceed the Vf. In the example of 100 2Vf LED's (IN SERIES) (200Vf total) (- in theory - keep that in mind -) I would not choose 400VDC to power the 100 LED's. 202VDC would IN THEORY be sufficient. But due to variations in manufactured LED's, some temperature concerns - etc. using 210VDC or even 220VDC would be sufficient. If I had a supply voltage of 210 and wanted a 20mA current and I wanted to use a resistor I would opt for a 500Ω resistor. It would be a big resistor because it would have to dissipate 4.2 watts. So in this instance I'd need a 5 watt resistor. I shudder to consider powering 200Vf from a 400VDC source. (double the Vf as afore mentioned)

Unwatched from here on.

 

hi tony,
If I read you correctly.?
The series resistor has only the overhead voltage across and at 20mA.
So 220V -200Vled, == 20V*0.02A = 0.4W, so a 1Watt 1k, resistor would be fine.
E

 

@ anyone who cares:
Vs (Supply Voltage) should be a few volts MINIMUM above Vf. I said "MINIMUM". That is to say you need "Enough" voltage to light the LED's AND employ a resistor to limit the current in some reasonable fashion. There ARE cases where using a resistor is less desirable than a CC source, but for the sake of argument about Vs - you need enough voltage to push the current through the LED's.

Now; what about Maximum voltage? IN THEORY there is no limit. You can "IN THEORY" power a single 2Vf LED with 100kVDC. You'll just need a hell of a resistor or a hell of a constant current source. You don't "Always" double the Vf. You merely exceed the Vf. In the example of 100 2Vf LED's (IN SERIES) (200Vf total) (- in theory - keep that in mind -) I would not choose 400VDC to power the 100 LED's. 202VDC would IN THEORY be sufficient. But due to variations in manufactured LED's, some temperature concerns - etc. using 210VDC or even 220VDC would be sufficient. If I had a supply voltage of 210 and wanted a 20mA current and I wanted to use a resistor I would opt for a 500Ω resistor. It would be a big resistor because it would have to dissipate 4.2 watts. So in this instance I'd need a 5 watt resistor. I shudder to consider powering 200Vf from a 400VDC source. (double the Vf as afore mentioned)

Unwatched from here on.

A couple of things that you need to consider. If the typical Vf of the diodes is 2 V, the maximum Vf is likely 2.5 V, meaning that you could need as much as 250 V just to get the targeted current flowing in them. Unless the parts were binned, it is unlikely that all of the parts will be that high, even if they all came from the same wafer, but there can be a significant systematic shift.

If you only allow for 10 V overhead above the nominal 200 V required, you are relying on those 100 LEDs to have an average Vf of no more than 2.1 V (and, at that point, you would have no remaining overhead voltage for your current-setting resistor to work with).

You are making the common rookie mistake of taking a voltage from one place (the supply voltage) and multiplying it my the current in order to get the power someplace else (the resistor). You need to use the voltage across the resistor, which is just 10 V. If you have 10 V across and 20 mA through, you only have 200 mW. The total power delivered by the supply would be 4.2 W, but only 200 mW would be dissipated in the resistor. The other 4 W would be dissipated in the LEDs at about 40 mW per device.

If I actually had to power a string of 100 LEDs, my first choice would be to use a switch-mode current source so that it would generate just the right amount of voltage to drive the LEDs at the desired current.

If that wasn't an option, my second choice would be to purchase binned LEDs (or bin them myself) to get the uncertainty in the Vf down as much as I could.

If that wasn't an option, and I had to make a lot of these, then I might consider purchasing all of the LEDs at once and, via sampling, estimate what the Vf distribution looks like and then randomize the selection of LEDs for each string so that they are drawn from that distribution.

If I were forced to bullet-proof the design on paper for any possible selection of LEDs, then I'm going to have to use a much higher overhead voltage. But I'm not going to WAG it or throw some rule-of-thumb at it (like I might if I were working with just a few LEDs).

I want Imin when the LEDs are all at Vfmax and I want Imax when the LEDs are all at Vfmin.

If the Vftyp is 2.0 V, the Vfmin is likely somewhere around 1.8 V. Most datasheets don't spec Vfmin, but those that do seem to be in that ballpark. I could consider ordering LEDs that have a Vfmin spec'ed.

I = (Vs - N⋅Vf) / R

I(Vfmin) ≤ Ifmax
I(Vfmax) ≥ Ifmin

This means that we have:

I(Vfmin) ≤ Ifmax
(Vs - N⋅Vfmin) / R ≤ Ifmax
(Vs - N⋅Vfmin) / Ifmax ≤ R

I(Vfmax) ≥ Ifmin
(Vs - N⋅Vfmax) / R ≥ Ifmin
(Vs - N⋅Vfmax) / Ifmin ≥ R

Combining these, we have

(Vs - N⋅Vfmin) / Ifmax ≤ R ≤ (Vs - N⋅Vfmmax) / Ifmin

Let's solve for the boundary case where the two limits are equal and there is a single resistor value.

(Vs - N⋅Vfmin) / Ifmax = R = (Vs - N⋅Vfmax) / Ifmin
(Vs - N⋅Vfmin) / Ifmax = (Vs - N⋅Vfmax) / Ifmin
(Vs - N⋅Vfmin) = (Vs - N⋅Vfmax) (Ifmax / Ifmin)
(Vs) - (N⋅Vfmin) = (Vs)(Ifmax / Ifmin) - (N⋅Vfmax) (Ifmax / Ifmin)
(N⋅Vfmax) (Ifmax / Ifmin) - (N⋅Vfmin) = (Vs)(Ifmax / Ifmin) - (Vs)
(N⋅Vfmax) (Ifmax / Ifmin) - (N⋅Vfmin) = (Vs)[(Ifmax / Ifmin) - 1]
Vs = N ⋅ (Vfmax⋅Ifmax - Vfmin⋅Ifmin) / (Ifmax - Ifmin)

Vs = (N⋅Vfmin) (Vfmax/Vfmin⋅Ifmax - Ifmin) / (Ifmax - Ifmin)
Vs = (N⋅Vfmin) [(Vfmax/Vfmin)⋅(Ifmax/Ifmin) - 1] / [(Ifmax/Ifmin) - 1]

For Vfmin = 1.8 V and Vfmax = 2.5 V and Ifmin = 10 mA and Ifmax = 30 mA, this works out to

Vs = 285 V

If Ifmax is lowered to 20 mA, then Vs goes to 320 V.

Note that this is not taking temperature effects into account, and it is also assuming no uncertainty in the supply voltage or the resistor value. Other than that, however, it is an extremely conservative design that makes no assumptions about the distribution of the LEDs except that they meet spec. The needed value of Vs can be brought down quite a bit if you are in a position to make some assumptions, either by knowing more about the distribution of the LEDs, or my doing board-level testing and choosing the value of R for each board, or by doing acceptable testing of each board.

 

A couple of things that you need to consider. If the typical Vf of the diodes is 2 V, the maximum Vf is likely 2.5 V, meaning that you could need as much as 250 V just to get the targeted current flowing in them.

I think @Tonyr1084 was speaking in general terms. None of us expect something to be exactly as per spec. Tony also eluded to

due to variations in manufactured LED's, some temperature concerns - etc.

Acknowledgement to variables was addressed.

From what I gathered, Tony was speaking theoretically, as he highlighted several times (sheesh!).

 

I think @Tonyr1084 was speaking in general terms. None of us expect something to be exactly as per spec. Tony also eluded to

Acknowledgement to variables was addressed.

From what I gathered, Tony was speaking theoretically, as he highlighted several times (sheesh!).

Go back and read it again.

In the example of 100 2Vf LED's (IN SERIES) (200Vf total) (- in theory - keep that in mind -) I would not choose 400VDC to power the 100 LED's. 202VDC would IN THEORY be sufficient.

That was the "in theory" part.

But due to variations in manufactured LED's, some temperature concerns - etc. using 210VDC or even 220VDC would be sufficient.

The claim is explicit. AFTER allowing for variations in manufacturing, a voltage of 210 VDC or 220 VDC would be sufficient to DEAL with that variation.

Furthermore, the post was in keeping with the general discussion of the thread, namely what supply voltage should be used, and not so much as responding narrowly to just that specific post.

 

NONE of us would ever expect to put 100 LED's in series. If not theory then use rational reasoning.

 

NONE of us would ever expect to put 100 LED's in series. If not theory then use rational reasoning.

Tell that to the people that make string lights powered by 230 VAC with a single blocking diode and current limiting resistor. Depending on color, those often have 100 (or more) LEDs in them.

It's the exact same rationale as series-wired Christmas tree lights. Yes, if one goes out, they all go out. Didn't stop them from making millions and millions of them that way.

 

The ones I've seen have no more than 13 LED's in a series set. Many of my strings (short ones) can lose one LED and only half go out.

Now, what decent engineer would build a circuit using that many series LED's?! I wouldn't. Would you?

 

The ones I've seen have no more than 13 LED's in a series set. Many of my strings (short ones) can lose one LED and only half go out.

Now, what decent engineer would build a circuit using that many series LED's?! I wouldn't. Would you?

Just because you wouldn't do something doesn't mean it isn't done.

As to whether I would do something like that, I would try to avoid it, but at the end of the day it depends on what the project was and what was and wasn't important to the customer.

 

Just because you wouldn't do something doesn't mean it isn't done.

YouTube is testament to that statement. Seen LOTS of stupid things done on YouTube.