resistance values for LED's

Hi Fanie,

Putting LEDs in series can certainly work to lesson the heating loss of the system, but I have not had good luck with that approach and had multiple failures when I tried it. I think the reason was because the LEDs I had were not well matched, even though they were from the same production batch. I theorized at the time that the failure mechanism I was experiencing was that in several incidents for some reason one LED in the string would have a higher resistance, it would heat up more, thus causing the already high resistance to increase, more voltage would then develop across it, and it would finally go into a "run away" mode and self-destruct --- thus killing the whole string. This was several years ago, but I think I tried it three times and each time the string failed. These were strings of 10 LEDs. This is where a constant-current source can cause problems that do not occur when only one LED is involved.

Your solution of putting diodes in series with the LEDs will not actually do anything that a simple resistor doesn't do, except block a reverse voltage of course. The forward voltage drop across anything (diode, wire, transistor, etc.) will result in the same heat loss at any particular current. This follows from the "Law of Conservation of Energy."

I like your circuits from post #57, especially the variable duty cycle PWM dimmer (the second circuit). Unfortunately, your dimmer, being an on-off switcher, will also produce high frequency transients and should be shielded or the frequency adjusted to a non-interfering frequency.

The problem with anything that uses fast switching is often not the fundamental frequency of the switched current, but the harmonics. If you put a "square wave" signal into a spectrum analyzer you will see the fundamental and then a series of third-harmonic frequencies that march off to ever higher frequencies until the slope of that highest frequency matches the slope of the rise and fall of the original "square wave" which is, of course, never actually instantaneous. Knowing that this is going to happen I agree with Will that you just design your system to deal with it and the advantages of switchers and PWM are then yours for the taking.

BillyDoc

 

Sigh. Because it is part of the crap I work with, designing electronic stuff. I'm not really full time into boating



Billy, LED's does not work with voltage, they work with current. I'm sure there was a small overlook hence the failure.

Re-read post #45

The two circuits will have ZERO emission other than light and ZERO transients and will cause ZERO interference.

The harmonics is generated by an inductor's switching and works like a little antenna transmitting all over the show. The two circuits I uploaded does not have any inductors, hence no rf interference can occur.

Build either circuits, they work very well.

I'll help you get it right.

 

This is the answer Fanie! No-voltage LEDs. It solves so many problems! But . . . where does the current come from?

Then I must not be understanding the second circuit. It looks like an oscillator, to me, with a variable duty cycle (the diodes and potentiometer) feeding three NAND gates in parallel (I assume the output stage of the NAND gates is MOSFET, not bipolar, or has a current limit of some sort) which will output a square-wave with variable duty cycle, thus dimming the LEDs. Am I misreading it?

Inductors are not the only source of rf emissions. Any conductor that has a variable current on it can transmit and does if not shielded. Consider an ordinary antenna. Here in the US we use 60 Hz power, and I can stick my oscilloscope lead in the air almost anywhere and pick 60 cycle up. Of course, as US citizens we are all quite accustomed to continually being fluxed. Magnetically and otherwise.

I don't doubt it! As I said, and meant, the second circuit is quite nice. Did you design it? But NAND gates have very fast rise and fall times, so they are going to also have rf in their output. Whether it interferes or not is another issue.

I don't doubt this either, and I truly appreciate the effort. I learn a lot from my mistakes and even take the philosophical position that when I'm NOT making mistakes I'm not trying hard enough.

BillyDoc

 

Led's are designed to work with a certain current flowing through them. If you take two 20mA LED's, then the voltage drop across them may have different values, but the current they require to full on is the same.

For instance, the LED's I will be using for my lights are rated for 350mA. I tested the light level and at 200mA it is slightly less than at 350mA, but the power consumption is only 60% of full current (almost half). The current source as well as the other components will hence be designed to accomodate the 200mA.

As you can see the voltage is of secondary interest. All you have to make sure of is that the supply voltage will not be lower than the total of the whole circuit, in which case you won't be able to maintain full on brightness.

That is about how the circuit works, it is not rocket science. Any one can see that

Any conducter with current flowing through it will develop an electrical field around it. If you switch the current on and off then the field will collapse and pop up with each switching. A conductor is not necessarily an inductor !

This magnetic field however is minute, and remember you have not wound a coil with it, and there are no resonances (harmonics) either, just pure switching on and off.

If you wind a coil with wire but no core, the permiability is something like 1.
If you stick a piece of iron in it it goes up by ~ x10.
Other materials ie ferrite cores or metal iron cores increase it even further.

(A switched coil wil charge and discharge itself as the energy is lost and it looks like a sine wave going smaller and smaller. If these resonances coinside with the next switch then you get all kinds of weird and wonderful things happening in the form of emission.)

If you, by the worst bad luck I ever heard of, design a tuned circuit with that PWM frequency and the wire length of the components used, then all you have to do is chage the pot value or the cap value to de-tune it. My guess is if you get that right you will be the first

All in all the current is so low that you are really going to have problems picking up those magnetic fields on the conductors.

I always design my own circuits. If it wasn't I'd send you a link.

It's ok to theorise but philosophy has nothing to do with electronics It's so easy to just to ask someone and they will help you, you don't have to learn from your own mistakes. If we all do that, half of us would have drowned by now

 

Hi Fanie,

I think we are largely in agreement on this issue, but there are a couple of things I would like to clarify.

First, on the LED light output. Your data imply one of two things, either the efficiency of the LED is changing and it is becoming more efficient at lower currents, or your perception of the light intensity is incorrect. I know nothing about light output efficiency as a function of current, but I do know that human luminance perception is notoriously poor. There are a couple of reasons for this. The dominating reason is that as light intensity increases your eyes compensate by adjusting your pupils to limit that light. This is a completely involuntary reflex and not brought to consciousness. Secondarily, there appears to be modest neural control of the perceived intensity somewhere in the path to the brain from retinal receptors. You can easily check your hypothesis (higher efficiency at lower working current) by setting up a photodiode or other photoreceptor and taking some measurements. If you are right, then that is a very useful bit of knowledge to have! In any case, I think that it is well established that lower currents increase the lifetime of the LED considerably. At my age though, I would rather use fewer LEDs for a given output. I'm planning on dying before they do anyway.

The second issue I would like to bring up concerns the spectra of square-wave-like electrical phenomena. According to a French mathematician by the name of Joseph Fourier, all such phenomena can be broken down into components which are completely comprised of sine and cosine waves. Fourier invented a mathematical technique for doing this, now called the Fourier transform, which was usefully elaborated by Culley and Tukey to the Fast Fourier Transform (FFT), which has become the foundational mathematical tool of all computer based spectrum analysis that I am aware of. Anyway, if you do a FFT of a 50% duty cycle square wave the result will be a fundamental sine at the same frequency, plus a third harmonic with a lower amplitude and another harmonic of that at a lower still amplitude and so on infinitely . . . if you are talking about a mathematically generated square wave. You can reverse the process and re-combine these sines and cosines to reconstruct the original as well. Anyway the point of all this is that ANY square-like wave, like PWM outputs, also contains lots of higher frequency components beyond the fundamental. And don't think these "higher frequencies" are mere mathematical abstractions, because they aren't. They physically exist within the electrical circuit and behave exactly as you would expect signals of those particular frequencies and amplitudes to behave. If you have access to a spectrum analyzer, put in a square wave and analyze it. You will plainly see the fundamental, plus all the harmonics as I describe.

This is why I say that your PWM circuit will act just like any other switch-mode circuit that is producing a fast rise and fall time waveform, and even if you are switching at a single Hz or less there will be higher frequency components that will radiate . . . unless you take precautions. Fortunately, as Will and I have both mentioned, these precautions are pretty easy to take, you just need to know to do them.

BillyDoc

 

Billy, there are electronic forums. Post on them and see if what I said is so or not.

 

If no one is going to attempt the circuit(s) then I have done all this for nothing

 

Thanks for that Billy Point is no matter how many times I say that the circuit I uploaded has no emission, you keep on arguing the point. Which is why I suggested you ask someone else. Maybe you will believe them.

I have no desire for a lengthy argument of is, is not, is, is not.

 

Relax, Fanie!

Who's "coped out?" I kept on arguing the point because I think you are wrong, and I don't like such misconceptions to propagate to others. By discussing it publicly any reader has the information he needs to draw his own conclusions. This stuff is magic for most people, remember.

I'm sorry you don't want a lengthy discussion, but you just threw down the gauntlet, didn't you! I was leaving it where it was.

You will note, I hope, that I pointed out above that there really isn't a problem if you understand what you are doing, even a little.

Here is the crux of the issue: You have argued that: "The harmonics is generated by an inductor's switching and works like a little antenna transmitting all over the show. The two circuits I uploaded does not have any inductors, hence no rf interference can occur." So, either you are taking the position that a simple linear wire, like a wire antenna for example, is an inductor because it most certainly can broadcast rf energy --- in which case so can your second circuit if you use wires to glue it together; or you are saying that the wires in your second circuit are NOT INDUCTORS (which seems to be the case) and the wires in your second circuit therefore can't broadcast rf energy. I think you see that this won't fly, given that a wire antenna works pretty well to broadcast rf energy.

You were also quite adamant about this issue in post #62, stating that: "The two circuits will have ZERO emission other than light and ZERO transients and will cause ZERO interference." And I'm very sorry, Fanie, but you are just flat-out wrong about this. I just checked the data sheet for the Fairchild version of the 4093 chip you call for (CD4093BC) and see that the transient times for the gates are 50 NANOseconds, and faster! That will have a LOT of radio frequency energy if, as your circuit indicates, you use those outputs to directly drive three LEDs. In your discussion in post #45 you talk of driving three LEDs at 100 mA, with a 9 volt drop. That's 0.9 watts of rf energy right there, not counting the rest of your circuit! And this energy will likely be in close proximity to a GPS which is trying to make sense of signals that are buried in noise under the best of conditions.

But I want everyone to be clear on one over-riding point: in my, not so humble, opinion, Fanie's second circuit is a nice bit of work and will most likely do the job it is intended to do. The issue of failures from mis-matched LEDs is minor because modern manufacturing processes match them quite well in most cases. You have to be pretty unlucky to get a sufficient mis-match to cause trouble. If you don't have issues with rf interference in you boat, no problem. BUT, if you do have issues then you might want to consider the following: when using this circuit put it in a metal box and use shielded cables to the lights, preferably twisted pair shielded cables, only ground the cable shields at one end (where they come together with the box if they are multiple) and if you want an absolutely radio frequency quiet installation put the LEDs behind some copper screening that is grounded in the same way as the cable shields (daisy-chained with the shields is OK). OR, just build this simple circuit into the same fixture as the LEDs and put a fat capacitor across the input (power) leads. Either way you get the power saving of PWM and your costs are a few cents for the shielded cables (perhaps) and some extra effort up front. I think it's worth this extra effort!

BillyDoc

 

Lets put this interference / emission thing to rest shall we.

The frequency the circuit gets switched at is about 120Hz, see the attached timing graph for RC's.

At 120Hz, even IF there was any emissions, it is no where near any equipments operating frequencies.

Furthermore, the current source has a resistor that is going to absorb the initial switching bump if the FET switches on fast. The 317 is good for up to around 1MHz if I remember right, so at 125Hz it is going to idle.

Also if you look at the on charactristics of the LED's, they do not have a dead on set voltage drop over them, this does vary with current. So they will also with a minor slope turn on, because at a low current they already start to conduct initial current.

Although this happens very fast in human terms, it is slow enough in electronics not to cause interference. The current at 200mA is so low, I really don't see how you are going to get emission out of the circuit.

Maybe you should build it and show me the emission levels (other than light) you are going to get from it. I need your testimony on it

Lastly, I have had LED equipment surveyed before, it is a test similar to CE, and although the purpose was different, the emission levels was ZERO. That is how I know.

If the circuit was switched using a micro processor, then there is a minute amount of measurable emission, but that is caused by the micro running in the MHz clock frequency and defenately not enough to cause trouble for other equipment in any case.

So I remain adamant about my claims.

 

Thank you Fanie, good idea! Let me answer your points one by one.

This would be quite true if you were powering the LEDs with a sine wave, but you are not. It's the harmonics from the square wave output that are the problem. Your circuit provides a variable duty cycle square wave output with rise and fall times of about 50 nanoSeconds from the NAND gates, and probably something nearly as fast from the FET. A 50 nanoSecond rise and fall time suggests a frequency of one divided by 100 nS, which is 10 MHz . . . the highest harmonic frequency you actually need to expect. A square wave with a 50-50 duty cycle will be comprised of a series of sines with frequencies that start at the fundamental (120 Hz in this case) and rise by thirds. That is, a series of harmonics at: 360 Hz, 1,080 Hz, 3,240 Hz, 9,720 Hz . . . and so on. I think that you can see that you can quickly get into a frequency range that can interfere because of this effect, up to the 10 MHz limit mentioned above. Moreover, the fact is your circuit is designed to provide a variable duty cycle . . . which means harmonics at a wide range of initial and subsequent harmonic frequencies that will be unknown for the most part. And speaking of unknown frequencies, just the capacitor value in your circuit usually varies by plus or minus 20%, making designing to a specific frequency quite difficult if you are trying to dodge specific interfering frequencies.

If you want to "absorb a bump in voltage" you need something that can store energy, like a capacitor or an inductor. A resistor is actually an infinitely fast device, limited by any inductance or capacitance it might posses. The 1MHz you mention is the response time of the device, meaning that it can react to a current change that fast but will be unable to react at faster speeds. In this case that means that it will supply current and track the changing conditions it is supplying right up to 1 MHz . . . so it will indeed follow the square wave slope with a rise and fall time of 0.5 microSecond, and the manufacturer guarantees it! It will actually do better in practice, but at a lower response amplitude. So this isn't going to filter anything out.

I think you will be hard pressed to find a LED that has a slow turn on and off. They are, after all, used to drive gigaHertz fiber optic networks and appear to be primarily resistive in nature. As you say, they do indeed turn on with less brightness at lower voltages and the brightness increases as the voltage does, but this does not in any way mean that they do this slowly!

I wish I could Fanie, and would do it gladly but I have no equipment with which to measure emission levels.

As you know,there are many ways to drive LED lamps and many ways to test. In the instance you cite, were the LEDs driven with fast-switched square waves? Were rf emissions measured at the same time?

As I have contended all along, the switching transients need not be a problem. They are in fact easy to deal with, if you know what you are doing! But, as Kim points out in post #39 you really ought to pay attention to the issue, not ignore it.

Me too!

BillyDoc

 

So you're here only to critisize me then, and make me look bad infront of every one else

My boards will arive in the next couple of days (you see, I'm not just all talk).

These will be uP controlled, but I will do a spectum scan on the emissions so we can see what emissions are from the LED PWM circuit. The clock will either be 4 or 8MHz but probably 8MHz so one would be able to distinuish between which is which. I'll make a special post for you on this, ok

You can help me with the code in the meantime, I'm using the baby pic 12F683.

How do I set the darn thing up ??? And which PWM frequency would be the best to use ?

 

Nah, Fanie, I don't think you "look bad in front of everyone else" at all, and doubt anyone else would either. We just have a difference of opinion over a technical issue, no big deal. I will be interested to hear your results from your new boards. Why do you need a microprocessor?

I'm not familiar with the PIC products at all, so can't help you with that. I built my first computer around a Zilog Z-80 when they first came out sometime in the '70s and have been using their products whenever I need a micro. Zilog provides a very nice IDE (Integrated Development Environment, free!), their prototyping boards, and the chips, are cheap and all programming is in ANSI C (or assembly) and I'm already familiar with their weird assembly language.

For the PWM frequency your 120 Hz is very good. This is because human vision will integrate flickering light sources at all comfortable luminances at a lower frequency still, so it will appear as a steady light. People who study these things (and that included me at one point in my career) call the phenomenon where the appearance of flicker transitions to the appearance of a steady light the "Flicker Fusion Frequency," and this varies by wavelength and luminance as well as frequency. I once published a paper demonstrating that the channels carrying visual information to the brain have different time constants for red and green light by phase shifting the two colors to a null flicker. Now I'd have to look up my own paper to tell you which one is faster! I think you see red faster than green by a few milliseconds, but that's just a guess at this point.

BillyDoc

 

Of course I look bad in front of everyone else. Nobody likes me any more. All they do is critisize by zero emission led circuits. Nobody agrees with me any more either. Not even the wife does

Zjeezzzz Jeff, can't we get a crying smiley as well eh, how is a man supposed to express his true feelings on this forum.

I don't know why I need the micro processor. I just like stuffing them into everything

What is a Z-80 ? Sounds like a Nissan or bee em trouble you. I'd be carefull using things with a Zee in the model. Kinda sounds like the end of the line.

The 120Hz is good ? Why ? The micro cannot run that slow so I'll probably rip the pwm up to 20 something kHz to stay out of the audible band and into the the radio frequencies. See this is where the micro is great. I can program it to sweep the freqiencies just so when it begins to be a problem for someone I change the frequency and it becomes someone else's problem, while through all this I have light and I stay out of trouble

 

Something tells me I'm older than you Fanie, have a little respect for your Seniors!

I suspect this because you didn't recognize the Z-80 chip, the fastest chip and best of breed (at the time) that DOUBLED the slow Intel 8080 chips clock speed of only 2 MHz to a screaming 4 MHz!!!! I ran CP/M on it, and loved it. You do remember CP/M, right? Ok! I even implemented memory bank switching by 2K segments and could address a full 256 K of memory when the 8080 struggled to address a mere 64 K of memory. Of course, it cost me some precious body parts to buy all that memory, but hey, what's a geek to do? I had the hottest box of anyone I knew . . . for a while. I used the thing for 10 years or so before I striped it for parts and it performed flawlessly all that time. Which is more than I can say for the computer I am writing this on.

The Z-80 was a classic chip(http://en.wikipedia.org/wiki/Zilog_Z80)! And Zilog gave Intel some serious heartburn when they brought it out. Check out their current micros at Zilog.com. Their support is excellent as well.

(Now where is that phone number to get my "reward" for mentioning them . . .)

BillyDoc

P.S. Fanie, NOBODY's wife agrees with them. It's in the rule book.