Custom led advice?!

Not to add fuel to this fire, but if you made 10 or 100 of these circuits and you knew the distribution of LED voltages at some current beforehand, you could predict

how many circuits would have some or all LEDs lit and

how many would have no LEDs lit and

how many would have LEDs lit and overdriven and so will not last long.

I doubt that the LED voltage fits a normal distribution and when you add in the resistor value tolerance you get very weird looking, non-normal distributions. Skewed, kurtosis, the whole bit.

It's a gamble and some people have a circuit like this working the first time - by luck, not by skill or knowledge.
 
Sponsored Links
Not to add fuel to this fire, but if you made 10 or 100 of these circuits and you knew the distribution of LED voltages at some current beforehand, you could predict
how many circuits would have some or all LEDs lit and
how many would have none LEDs lit and
how many would have LEDs lit and overdriven and so will not last long.
I doubt that the LED voltage fits a normal distribution and when you add in the resistor value tolerance you get very weird looking, non-normal distributions.
Indeed one could, but that would only really be of interest to a very bad designer, or as a classroom exercise. A good designer would choose supply voltage and resistor values such all circuits had all LEDs lit to roughly the right brightness and none would be 'overdriven'. In practice, it's really very easy to achieve that.

As I just wrote to BAS, the greater the difference between supply voltage and LED Vf, the closer one gets to a situation in which current is independent of supply voltage and variations in individual LED characteristics.

Kind Regards, John
 
Indeed one could, but that would only really be of interest to a very bad designer, or as a classroom exercise. A good designer would choose supply voltage and resistor values such all circuits had all LEDs lit to roughly the right brightness and none would be 'overdriven'. In practice, it's really very easy to achieve that.

Try it.
Assume the LED voltages at 20 mA are approximated by 3.5, 3.5, 4 and 2.5v (the unspecified low end of LED voltages). With three LEDs this is 4x3 values.
Then add voltage distribution, assume 12.8, 13.5, 14.4, 14.4, 15.
Then add resistor tolerance, assume a normal distribution with only three values - the center value, + 5% and - 5%.
Plug these values into the circuit and look at your 12x5x3 values of current on a spreadsheet, then rank them by order of likelihood.

For more accuracy, approximate the distributions with more values more nearly approximating the real-world distribution of values, if you can find this info.

Some things are out of the hands of the best designers.
A lot of managers overconstrain a problem and then wonder why their people are having trouble.
I once told the company president that their new product will not meet the published specifications.
When a consultant in the room saw where this was going he just kept repeating, "Everything will be all right."
The company president did not object to this magic chant; in fact he maybe uses it himself to get to sleep at night.
And these guys are supposed to be scientists, engineers, rational people, etc.
I'm surprised they didn't put on a Jamaican accent and wave chicken bones at me or make a little doll that looked like me and then stick pins in it.
 
Try it. Assume the LED voltages at 20 mA are approximated by 3.5, 3.5, 4 and 2.5v (the unspecified low end of LED voltages). With three LEDs this is 4x3 values.
Then add voltage distribution, assume 12.8, 13.5, 14.4, 14.4, 15.
Then add resistor tolerance, assume a normal distribution with only three values - the center value, + 5% and - 5%.
Plug these values into the circuit and look at your 12x5x3 values of current on a spreadsheet, then rank them by order of likelihood. ... Some things are out of the hands of the best designers.
I don't need to actually do it to understand your point but, in practice, if one doesn't like the answers, one's only options are to specify tighter tolerance components and/or to tweak the value of the resistor in each case to achieve what one wants. Also, as I said before, the higher the supply voltage, the less effect will result from variations in LED Vfs.

In reality, particular if one sources from decent manufacturers, the operating characteristics of LEDs and the resistances of resistors are usually considerably closer to their nominal values that the specification limits may suggest - so that, in practice, there really is not usually much of a problem. I've certainly never had a problem with driving LEDs, and I hate to think how many I've used over the years/decades!

Kind Regards, John
 
Sponsored Links
The hour is late, but having fortified myself with milk, sugar and bread I now realize that your use of qualifiers is beginning to sound like this http://en.wikipedia.org/wiki/No_true_Scotsman fallacy. :D
It will not surprise you to hear that I don't see it like that!

I have perhaps not been making my points as clearly as they exist in my mind :) The set of figures you asked me to 'try' will give very 'unsatisfactory' results mainly because they violate one of the principles I have repeatedly mentioned throughout this discussion, and hence would represent 'bad design' - that principle being that, to achieve a reasonable degree of control over LED current (and, by implication, 'reliability of design') one's lowest possible supply voltage needs to be 'appreciably' (ideally quite a long way) above the highest possible Vf of the LED(s).

The 'worst case' with your figures has Vf=4V*3 = 12V and supply voltage of only 12.8V, a difference (across the resistor) which is far too small for reasonable control over LED current. Hence, if one knew that one's supply voltage might fall as low as 12.8V, one should not even be thinking of trying to run LEDs which could have a total Vf of 12V from such a supply, since that would be 'very bad design'. If, as Bernard and I have suggested, you changed it to just two LEDs (i.e. 'a far better design'), so that the lowest voltage across the resistor (in the 'worst case') rose from 0.8V to 4.8V, then your spreadsheet would look much happier! Drop it down further, to just one LED, and your spreadsheet will be even happier :)

Kind Regards, John
 
Assume the LED voltages at 20 mA are approximated by 3.5, 3.5, 4 and 2.5v (the unspecified low end of LED voltages).
That range of Vf from one type of LED with a set current is very un-likely to be found in practise other than is some very poorly manufactured devices. If a batch of components showed that range on test then they would ( should ) be rejected as being of dubious quality. In the early days of LEDs they were often graded by brightness testing by either the manufacturer or the user's goods in quality control


Some things are out of the hands of the best designers.
Designers need to be able to promote their design against any fatal constraints imposed on the design. I have often had the design to market budget increased by showing the cost savings in future servicing and maintainance of the item being designed. Getting constraints reduced to increase long term profit from the item is part of the good designer's remit
 
Assume the LED voltages at 20 mA are approximated by 3.5, 3.5, 4 and 2.5v (the unspecified low end of LED voltages).
That range of Vf from one type of LED with a set current is very un-likely to be found in practise other than is some very poorly manufactured devices.
Indeed so. As I wrote:
In reality, particular if one sources from decent manufacturers, the operating characteristics of LEDs and the resistances of resistors are usually considerably closer to their nominal values that the specification limits may suggest - so that, in practice, there really is not usually much of a problem.
As I implied, I think one problem is that people often misinterpret specification limits, thinking that they are a direct indication of the production spread (e.g. '2 Standard Deviations from mean'). Although that may sometimes be true, they are more commonly much wider than that (i.e.'many SDs from mean') - such that 'virtually no' products will be outside of those limits. In that situation, the great majority of products will close to nominal values and nowhere near the 'limit'.

As I wrote to Porque last night, IF, for whatever reason, one were constrained to work with LEDs having as wide a scatter of characteristics as he suggested, and IF he wanted it to work over the range of supply voltages he indicated, then his statistical calculations would, indeed, indicate a significant probability of LED current ending up above or below an acceptable range. However, as I said, for that very reason, no competent designer would even propose such a design, since its inadequacy would be apparent without needing to do any of the calculations. Instead, (s)he almost certainly would (should!) increase the minimum possible voltage across the resistor by (as you suggested way back) reducing the number of LEDs from three to two. Porque could then do his statistical calculations which would hopefully show only a low probability of LED current falling outside of the acceptable range, even in the 'worst case' combinations of variation in supply voltage and component characteristics.

Kind Regards, John
 
Try it. Assume the LED voltages at 20 mA are approximated by 3.5, 3.5, 4 and 2.5v (the unspecified low end of LED voltages). With three LEDs this is 4x3 values.
Then add voltage distribution, assume 12.8, 13.5, 14.4, 14.4, 15.
Then add resistor tolerance, assume a normal distribution with only three values - the center value, + 5% and - 5%.
Plug these values into the circuit and look at your 12x5x3 values of current on a spreadsheet, then rank them by order of likelihood.
A designer doesn’t really need to look in detail at distributions to realise that (with the extremely wide range of LED characteristics you are postulating and the range of voltages you want to accommodate), a design involving three of the LEDs in series is a non-starter – one only needs to look at the ‘worst cases’ ...

For the purpose of calculation, I have assumed that the intention was to get an LED current of 20mA when the LED Vf was the mean of the figures you postulate (3.375V) and the supply voltage is the mean of the figures you quote (14.02V). That equates to a resistor value of 194.75Ω. I will therefore assume that you would select the neared E24 value (200Ω) and, for the purpose of illustration, further assume that all such resistors have values within 5% of that nominal value (i.e. 190Ω-200Ω).

The worst cases are then:
(a)...Vf = 4V*3 = 12V supply = 12.8V resistor = 210Ω, which gives an LED current of 3.8 mA, and
(b)...Vf =2.5V*3 = 7.5V supply = 15V resistor = 190Ω, which gives an LED current of 39.5 mA.

Both of those extremes are outside of an acceptable range for a ’20 mA LED’. Examining the distribution will indicate what proportion would be outside of the acceptable range, but the designer obviously should design such that the probability of any being outside of that range is very low.

In contrast, if one adopts a more sensible design using just two LEDs in series, the resistor value would then become 363.5Ω – so use a 360Ω E24 component with an assumed range of possible values of 342Ω-378Ω.

The worst cases are then:
(a)...Vf = 4V*2 = 8V supply = 12.8V resistor = 378Ω, which gives an LED current of 12.7 mA, and
(b)...Vf =2.5V*2 = 5.0V supply = 15V resistor = 342Ω, which gives an LED current of 29.2 mA.

Those extreme values are probably just about acceptable for a ’20 mA LED’. If one worked with a more realistic estimate of variability of LED Vf, the range of currents would certainly be acceptable. Examining the actual distribution of LED currents would add relatively little to this situation.

Having said that, I’ll do some simulations of the distributions and show you shortly what they look like in the two cases.

Kind Regards, John
 
Vs R LED v I
6 125 2.5 0.028
6 125 3.5 0.02
6 125 3.5 0.02
6 125 4 0.016

12 425 2.5 0.022
12 425 3.5 0.020
12 425 3.5 0.020
12 425 4 0.019

Assuming the only variable is LED voltage, if 16 to 28 mA gives an unacceptable variation in LED brightness, you could go to 12v if you have that option.

If you add in resistor tolerance, it's worse. If you add in voltage variation, it's worse.

If you build these circuits you have a 50% chance of getting 20 mA and a 50% chance of getting the other values.
 
Having said that, I’ll do some simulations of the distributions and show you shortly what they look like in the two cases.
OK, rather than work just with the few discrete values that Porque threw at us, I’ve attempted to look more generally. I have assumed that both the LED Vf and the supply voltage are Normally distributed, with a mean equal to the mean of Porque’s figures and with a variability (Standard Deviation) such that only about 1% of cases would be more extreme (further from the mean) than the figures Porque presented. For the series resistor, I have assumed the preferred value per my previous posting, and have assumed that values are Normally distributed around that value such that about 95% of values fall within ±5% of the nominal value.

I have then undertaken simulations of one million instances, in each case drawing the supply voltage, resistor value and Vf for each of the LEDs from the assumed Normal distributions described above. For three LEDs in series, the distribution of the resultant LED currents is:
With (‘better design’) two LEDs in series, the distribution looks like:
I think these are fairly self-explanatory, but would welcome any comments or questions. Looking at one million simulations drawn from assumed distributions is obvioulsy likely to give a clearer picture of what's going on than one gets by playing with a handful of discrete values in a spreadsheet. Needless to say, with more realistic figures for the variation in LED Vf, the distributions would be much ‘narrower’.

Kind Regards, John
 
<some example data, with one LED plus resistor fed from 6V and 12V>
Assuming the only variable is LED voltage, if 16 to 28 mA gives an unacceptable variation in LED brightness, you could go to 12v if you have that option.
I would personally probably regard that range of currents as acceptable but, as you say (and as I keep saying), the greater you make supply voltage relative to the LED's Vf, the less variation in current will you see due to variations in that Vf.
If you add in resistor tolerance, it's worse. If you add in voltage variation, it's worse.
Indeed - see my simulations, which include the variabilty of everything.
If you build these circuits you have a 50% chance of getting 20 mA and a 50% chance of getting the other values.
True - but, in practice, interpretation of that obviously depends crucially on what those 'othervalues' are, and whether or not they are acceptable.

Kind Regards, John
 
Porque

These diode voltages you are quoting....? are they for LEDs from the same batch or are they LEDs of varying colours ?.
 
Porque These diode voltages you are quoting....? are they for LEDs from the same batch or are they LEDs of varying colours ?.
Very good question. I obviously may be wrong, but I rather suspect that the 3.5V and 4.0V figures are taken from the data sheet that was linked (being the 'Typical' and 'Maximum' figures in that data sheet, respectively) and, from what he wrote, the 2.5V was the generically lowest that you're likely to see for any LED. If that's true, then the figures obviously bear no real resemblence to the variability one would expect to see within a batch, or even batches of the same component - the variability of which I would expect to be far lower than his figures imply.

Kind Regards, John
 

DIYnot Local

Staff member

If you need to find a tradesperson to get your job done, please try our local search below, or if you are doing it yourself you can find suppliers local to you.

Select the supplier or trade you require, enter your location to begin your search.


Are you a trade or supplier? You can create your listing free at DIYnot Local

 
Sponsored Links
Back
Top