What do I need to run 200 LEDs on 110V

[kevin071572]

I want to build my own under cabinet LED strips for my kitchen. I am looking at wiring 200 Nichia T1 3/4 LEDs to one AC circuit. But I am having problems finding any directions to build the circuit or driver to power a system like this. If I need to buy a premade step down transformer then I could to that too, but I don't know if that is the best way or not. Can somebody help?


Kevin

 

[russ_watters]

What are the power requirements of these LEDs?

 

[mgb_phys]

Is it practical to run LEDs from AC?
Presumably any flicker at 60Hz won't be visible- but what does it do to the life of the diodes?

 

[russ_watters]

I'm thinking no, but then a store-bought dc power source is pretty cheap...

 

[Eidos]

Ok cool, the one thing you want to do is to limit the current flowing the LED since you can break them if its too high, done this a couple of times myself

As has been said its easiest to go with a DC supply since AC won't be visible at 60Hz (if you're in the states) since the LED will flicker at about this frequency and our eyes can't refresh that quickly.

I found a cute web app that might help you out

http://led.linear1.org/led.wiz

Looks like someone has done all the work for us :)

 

[kevin071572]

I forgot about the possible flicker from AC, I knew I came to the right spot to ask about this. :)
So, I need to drop the voltage and convert to DC. That's easily solved. But my next question is what would be the ideal voltage to run this many LED's? If I bought a 110VAC - 12VDC supply, would that be easy to work with and reliable?

Eidos- THAT LINK IS PERFECT! THANKS

 

[NoTime]

The calculator above does not account for forward resistance,
Not an issue for one or two LEDs but for long strings of LEDs this changes.
However, worse case is they will be dimmer than expected.

You might notice flicker since they would only conduct for a half cycle.
I'm fairly susceptible to flicker and haven't noticed any for one off 1/2 cycle AC use.
But then I think wide area bright illumination is a factor in the flicker effect.

Adding a full wave bridge should fix that as well as eliminating any problem exceeding the reverse bias spec which is the only potential problem I can think of for AC use.
Leakage variances in string mode can easily cause the reverse breakdown spec to be exceeded.

If you use a ground fault and understand proper insulation techniques then, with 200 LEDs, they can be series connected for direct AC line use.
Otherwise a stepdown transformer is recommended.

Improper wiring/insulation can still be a deadly combination even with a ground fault.
Particularly with the long strings needed for this application.

 

[mgb_phys]

The typical way to do this would be to use a 12V dc supply.
Then hook up the appropriate number ( probably 8-10) LEDs in series to give a forward voltage drop of 12V, you might have to experiment slightly adding/subtracting a few to get the best brightness.
Then connect sets of these bars in parallel, checking the total current output of the 12Vdc supply.

 

[Averagesupernova]

You will still want some sort of series resistor even if you series up about 6 LEDs. At 2 volts drop each 6 LEDs will get you 12 volts. But the amount each one drops is dependent on temperature and a few other things. If you hook up 6 LEDs that each are dropping 1.8 volts to a 12 volt supply you will most likely fry at least one of them if there is no series resistor installed.

 

[Redbelly98]

If these are white LED's, each one has about 3 or 4 V drop. Not the 2V typical of red and yellow LED's.

Also, one should not assume a "12V" supply outputs 12V. It should be measured, then double-checked when running a load.

 

[kevin071572]

NoTime-If I were building these in a sealed enclosure then I'd look at running them at line voltage. Since this set up is going under cabinets, I am figured stepping down the voltage would safest, I just didn't know if I should convert to DC or what the best voltage would be for a circuit this large.

Would it be wrong to just run them all in parallel and use 1 resistor per 3 LEDs? That is what comes up when i enter the numbers on the link Eidos posted earlier. To get that I entered the following:
12V Source voltage
3.6 diode forward voltage
30 diode forward current (mA)
200 number of LEDs in your array

Thanks for all your help so far, this will be a fun project for a newbie like me.

 

[NoTime]

If you use a real 12v supply then you may not need the resistors.

Since there are several types of white LED and I haven't tried then out, I'm not sure what the forward resistance is on the ones you have.

The common red LED has a forward resistance of about 22 ohms.
If the white ones you have are similar then that would give you a built in 66 ohm resistor for three in series.
So you would not need to use the 47 ohm resistor which is less than 66 ohms.
Which is why I said the calculator need some work.

However, If 30mA is the max current rating for your LEDs then you should select for a lower operating current.
Running things at maximum specified ratings tends to fry them as the max rating also has a maximum operating temperature associated with it.

Note: You might be able to scavenge a nice 12v wall wart switching supply from a computer peripheral.

 

[Redbelly98]

Would it be wrong to just run them all in parallel and use 1 resistor per 3 LEDs?

That is the idea. The 12V power supply must be capable of supplying just over 2 Amps, if you are going to run 30 mA through all 67 strings.


200 LED's and 67 resistors will take a while to wire together. It's a really good idea to be sure the circuit will work before you go to all this trouble.

So:

Before you wire up all 200 LED's, wire up a single 3-LED + resistor string. In parallel with that, connect a resistor that will draw roughly 2 Amps, in order to simulate the total load of the complete 200-LED string. Six 1-ohm, 5-Watt resistors (in series to get 6 ohms) would suffice for this.

Double check that the LED's actually use the current you want for the resistor value you choose -- just measure V across the series resistor and apply ohm's law to find current. If necessary, adjust R (the resistor in series with the LED's) until you have the current you want.


Once you have experimentally determined R correctly, you can wire up the entire 200 LED circuit.

Good luck and have fun!

Mark

 

[Phrak]

ACH --- Fuse --- R1 --- C1 --- ~ + -----|------------A
....... Bridge...C2...Diode Strings (~ 50 to 100 volts per string)
ACN -------------------------- ~ - -----|------------C

R1 - negative temperature coefficient thermistor

 

[Phrak]

Is it practical to run LEDs from AC?
Presumably any flicker at 60Hz won't be visible- but what does it do to the life of the diodes?

You want to run the average current below specifications or the diode degrades in intensity and the lifetime is reduced. To be more refined, for any given duty cycle and frequency there is a maximum rated current. The higher the frequency and the lower the duty cycle the higher the current that you can run.

There's actually a net gain in running AC or repeated-pulse current for most LEDs (but not the white output variety). The eye sees the peak output, so your LEDs will actually appear brighter for the same average current.

 

[kevin071572]

You guys are just what I needed. Good honest, knowledgeable answers from people NOT trying to sell something.

Mark- That is so simple, it's no wonder I forgot about it. I've be researching and trying to find something to set up 200 LEDs, when I should have been trying to find out how to hook up a few and copy that setup to get my 200.

Phrak, NoTime-Your both right, I missed the column header that said "Absolute Maximum Rating" when I put in 30mA. When I looked lower on the spec sheet I see it say 20mA is typical for these LEDs.

NoTime, that's another good point about the internal resistance of the LEDs. I don't see anything on the spec sheet, but I'll test them when they come in.


Below is the link to the spec sheet, I haven't ordered them yet, but now that you guys have helped me figure this one out, I will defiently be ordering them and testing (read:playing) around with them this week.

http://www.nichia.com/specification/led_lamp/NSPW500CS-E.pdf"

Thanks again for all the help and I'll post the pictures after it's all done and running.
Kevin

 

[NoTime]

ACH --- Fuse --- R1 --- C1 --- ~ + -----|------------A
....... Bridge...C2...Diode Strings (~ 50 to 100 volts per string)
ACN -------------------------- ~ - -----|------------C

R1 - negative temperature coefficient thermistor

I'm curious about the point of C1.
That doesn't make any sense to me.

C2 isn't necessary since LEDs usually just care about RMS current.
It will just make the math harder, unless there is something special about the white variety.
It will also tend to eliminate the pulse mode effective brightness thing you mention.

Not sure about the point of R1 either.
Why the thermistor?
If resistors are required they should be in line with the LED series strings, since there will always be the threshold voltage variance issue.

IIRC only certain types of white LED have the current lifetime issue others do not.

 

[Phrak]

I'm curious about the point of C1.
That doesn't make any sense to me.

C2 isn't necessary since LEDs usually just care about RMS current.
It will just make the math harder, unless there is something special about the white variety.
It will also tend to eliminate the pulse mode effective brightness thing you mention.

Not sure about the point of R1 either.
Why the thermistor?
If resistors are required they should be in line with the LED series strings, since there will always be the threshold voltage variance issue.

Start with a simpler circuit; C1 in series with two antiparallel zeners, all across 120VAC. C1 limits the average current without resistive loss. Now, insert a bridge on the downstream side of the capacitor to get DC current to the zener. It has to be downstream for the DC current in the capacitor to be zero.

I don't know about C2 either. I threw it into be complete. It could increase LED lifetime a little.

There's a nice story about the thermistor. Originally, I gave this general architecture to a college, and forgot about it. He designed it in, sold it, and they blew up in the field! The thermistor is required to limit surge at power-up, should the circuit be turned on near an AC peak.

IIRC only certain types of white LED have the current lifetime issue others do not.

I'm sure you know more about LEDs than I do. I was speaking toward thermal degradation over time. Is there actually a total-current lifetime for white LEDs?

 

[OmCheeto]

... I haven't ordered them yet, but now that you guys have helped me figure this one out, I will defiently be ordering them and testing (read:playing) around with them this week.

Have you looked into just buying a strand of white-LED xmas lights?
I got a strand of 70 lights from Ace Hardware for $5 the day after xmas a couple years back.
You just plug them into your 110. No fuss, no muss.

Actually, I bought every set the store had. Then I raced off to the store about 4 miles away, only to see some geek nerd with an arm load of them headed for the check stand. Fortunately he couldn't carry them all so I got two more sets.

My entire house is lit with them now. And I have enough left over to make navigation lights for my boat, and brake and turn signals for my boat trailer.

Also, from my experience, LED wiring is best done by trial and error. I use a proto-board and start with the "correct" number of LED's, based on the calculations. Then I add and subtract bulbs until the light intensity vs. current flow reaches what I consider to be optimal levels. For instance, with 7 or 8 bulbs, the light level remains the same, but the current flow is less with 8 bulbs. Adding a 9th bulb to the series reduces the current level to the point where the light intensity is insufficient. But you also have to monitor the current flow in the circuit. Lowering the number of bulbs will make them brighter, but you don't want to exceed their maximum current capacity, or you'll have burned out bulbs.

 

[NoTime]

Start with a simpler circuit; C1 in series with two antiparallel zeners, all across 120VAC. C1 limits the average current without resistive loss. Now, insert a bridge on the downstream side of the capacitor to get DC current to the zener. It has to be downstream for the DC current in the capacitor to be zero.

I don't know about C2 either. I threw it into be complete. It could increase LED lifetime a little.

There's a nice story about the thermistor. Originally, I gave this general architecture to a college, and forgot about it. He designed it in, sold it, and they blew up in the field! The thermistor is required to limit surge at power-up, should the circuit be turned on near an AC peak.



I'm sure you know more about LEDs than I do. I was speaking toward thermal degradation over time. Is there actually a total-current lifetime for white LEDs?

C1 isn't necessary then.
The forward voltage drop with a string of about 32 white LEDs (3.6v) is sufficient.
Adding it just makes the math more complex and can cause issues with forward drop variance.
Without C2 you can put one downstream of the bridge.

R1 -- ok, but it is only potentially necessary if C2 is present.
I say potentially because other circuit parameters can perform the current limiting function it provides. You need to do a full circuit analysis.
It seems unlikely to help lifetime issues since peak current per the data sheet kevin071572 listed is 100ma.
Peak current with a 20ma RMS current is only about 28ma and heating is dependent on the RMS value.

I don't have a lot of experience with the white ones, but yes my understanding is that the problem with the specified LEDs would be thermal degradation of the phosphors.
I think the OLED variety has some direct issues with current as opposed to thermal effects, but I'm not entirely certain of this.

 

[NoTime]

Have you looked into just buying a strand of white-LED xmas lights?

I'll have to keep that in mind

 

[Phrak]

C1 isn't necessary then.
The forward voltage drop with a string of about 32 white LEDs (3.6v) is sufficient.
Adding it just makes the math more complex and can cause issues with forward drop variance.
Without C2 you can put one downstream of the bridge.

R1 -- ok, but it is only potentially necessary if C2 is present.
I say potentially because other circuit parameters can perform the current limiting function it provides. You need to do a full circuit analysis.
It seems unlikely to help lifetime issues since peak current per the data sheet kevin071572 listed is 100ma.
Peak current with a 20ma RMS current is only about 28ma and heating is dependent on the RMS value.

I don't have a lot of experience with the white ones, but yes my understanding is that the problem with the specified LEDs would be thermal degradation of the phosphors.
I think the OLED variety has some direct issues with current as opposed to thermal effects, but I'm not entirely certain of this.

I'm thinking you're right. Let's say you would want to run about 200 white LEDs with no more than 2.5 watts total loss in series reisistors (could be no resistance). Use worst case continuous AC high line of 110%, and 115% for 60 cycles. Did anyone link a data sheet? What's the dynamic resistance of these white variety LEDs? And the forward drop; white LEDs vary greatly from mfg to mfg.

 

[NoTime]

The OP posted a datasheet, last post first page.
From the Forward Voltage vs. Forward Current graph the dynamic resistance works out to about 22 ohms at 20ma increasing for higher currents.
So the BOE minimal circuit looks like
120v AC -- 0.1A fuse--diode (for potential reverse breakdown issues) -- 25 Leds.
RMS current about 20 ma. Peak current about 70ma.

Personally, if I wanted illumination then pulse mode looks like a good option.
Looks like you can increase the apparent brightness by a factor of 3 (edit: over DC).

 

[Phrak]

{Wait, stop the presses! I've had a brainstorm. A single component solution @ 19.95 bucks is available in the lighting section of your local home despot.}

What does BOE mean?

I don't like the data sheet. I'd guess 70 ma peak would thermally degrade the diodes. 60 Hz is so slow it's nearly DC for the thermal mass of the die embedded in plastic, wouldn't you say? The data sheet is just not good enough to tell us. I'd use a bridge, though. For the cost of three more diodes, the average current is doubled for the same peak current. I was thinking 44 series diodes and a bridge.

But then again, maybe you're right aiming at 70ma peak. Without better data or prior expertise, experimenting is in order.

 

[Phrak]

I couldn't think of the name. I'd like to see the family of thermal response curves. But that's not going to happen.

 

[NoTime]

{Wait, stop the presses! I've had a brainstorm. A single component solution @ 19.95 bucks is available in the lighting section of your local home despot.}

What does BOE mean?

I don't like the data sheet. I'd guess 70 ma peak would thermally degrade the diodes. 60 Hz is so slow it's nearly DC for the thermal mass of the die embedded in plastic, wouldn't you say? The data sheet is just not good enough to tell us. I'd use a bridge, though. For the cost of three more diodes, the average current is doubled for the same peak current. I was thinking 44 series diodes and a bridge.

But then again, maybe you're right aiming at 70ma peak. Without better data or prior expertise, experimenting is in order.

Data sheet looks fine to me.
It has all the information you need to work this out, including any thermal effects.

BOE = Back of Envelope. Quick and dirty but close.
Using a bridge bumps the freq to 120Hz, but doesn't add that many LEDs to the string.
I thinks it's around 28 or so.
Don't forget you're effectively lowering the supply voltage + adding resistance.
Peak current drops quite a bit.

There shouldn't be an issue with 70ma peak thermal since the actual duty cycle is low.
I also took worse case so the actual peak will be lower.
No idea if if actually would flicker with the single diode.
Experience says some people might see it and some wouldn't.

Homework assignment: Show the math for working the bridge case out
The answer is definitely not a count of 44.

 

[kevin071572]

Have you looked into just buying a strand of white-LED xmas lights?

OmCheeto-Great Idea, but this is the off-season and Ebay and Amazon don't have good enough deals on LEDs from christmas lights. Most of the strings I seen are approximately $1/LED.

Phrak, NoTime-You both were complaining about the data sheet, is there a better brand of LEDs besides the Nichia's? I'm not rigid about the brand, I just want decent LEDs.

NoTime-what did you mean about the home depot single component solution for $19.95?

 

[OmCheeto]

OmCheeto-Great Idea, but this is the off-season and Ebay and Amazon don't have good enough deals on LEDs from christmas lights. Most of the strings I seen are approximately $1/LED.

I found the box the lamps came in. The manufacturer, http://www.neo-neon.com/default.asp", apparently no longer makes the Christmas light strings. They appear to have expanded into just about everything else though. They even have a large selection of LED Disco Balls.

Moddersmart has white LED's in various sizes for $0.25 a piece. Loose of course.
The different sizes have luminous intensities varying by a factor of 10.
This is another parameter you'll want to consider. A 20,000 millicandela(mcd) lamp reflecting off of a tile countertop may be uncomfortable to look at.

 

[Redbelly98]

Phrak, NoTime-You both were complaining about the data sheet, is there a better brand of LEDs besides the Nichia's? I'm not rigid about the brand, I just want decent LEDs.

They were discussing running the LED's from 120 V AC, and concerns about peak current associated with AC operation.

As long as you still plan to run from a DC supply, those LED's will work just fine Have you gotten a price from a supplier?

 

[NoTime]

NoTime-what did you mean about the home depot single component solution for $19.95?

That was Phrak, but if what I saw when I was in there the other day is what Phrak is referring to, I think they are incandescent light ropes. Not sure. I didn't look that close.

PS:I don't have any issue with the LED datasheet.

 

[Phrak]

Data sheet looks fine to me.
It has all the information you need to work this out, including any thermal effects.
...There shouldn't be an issue with 70ma peak thermal since the actual duty cycle is low.

This is why the data sheet isn't fine. Say we drive 70 mA peak at 1/10 Hertz. There will be a long period where the current is over 30 mA. So how low a frequency is too low? The data sheet doesn't say. It provides no thermal response curves to answer the question: how much average current can be tollerated at a given duty cycle and given frequency without overheating the die at the end of each cycle?

Theramal response curves are readily available for transistors. Is there something I don't know about diodes?

 

[Redbelly98]

I thought Kevin had decided to go with a 12 V DC supply, in which case none of that matters. This stuff about peak currents and duty cycles is just a side-discussion between Phrak and NoTime, not relevant to what Kevin actually plans to build.

Please correct me if that is wrong.

 

[Phrak]

That's true Redbelly. Your right, I think, that omCheeto has the right idea. Experiment with diode count, and buy a premade product. Reverse engineer it if you like.

LEDs are too nonlinear for simple analysis (roughly exponential in current vs. voltage), another reason that explains our wide disagreement. Spice would be nice.

 

[Phrak]

Phrak, NoTime-You both were complaining about the data sheet, is there a better brand of LEDs besides the Nichia's? I'm not rigid about the brand, I just want decent LEDs.

That was me complaining, but just about the data sheet, not the LEDs. When you go to use something in a nonstandard manner, like driving LEDs with 60Hz AC, you don't always get the information you want.

NoTime-what did you mean about the home depot single component solution for $19.95?

That was actaully me. It was just a joke (buy a fluorescent under-the-counter lamp.)

 

[NoTime]

This is why the data sheet isn't fine. Say we drive 70 mA peak at 1/10 Hertz. There will be a long period where the current is over 30 mA. So how low a frequency is too low? The data sheet doesn't say. It provides no thermal response curves to answer the question: how much average current can be tollerated at a given duty cycle and given frequency without overheating the die at the end of each cycle?

Theramal response curves are readily available for transistors. Is there something I don't know about diodes?

Try looking at Page 8.
There is enough information there to work this out.

If you can't tell me where you got a count of 44 from, then I don't see any point to continuing this conversation.

 

[NoTime]

I thought Kevin had decided to go with a 12 V DC supply, in which case none of that matters. This stuff about peak currents and duty cycles is just a side-discussion between Phrak and NoTime, not relevant to what Kevin actually plans to build.

Please correct me if that is wrong.

You are correct, it is not particularly relevant to Kevin's current plans.

 

[Redbelly98]

Mind you, I have nothing against side-discussions and I think they're partly why Physics Forums is such a fantastic place. But I did want to clarify that for Kevin's sake.

At any rate, the main choices now seem to be either:
1) build a 12V DC LED string, or
2) find something pre-made if available, for less money and takes less time than making one's own.

 

[kevin071572]

Try looking at Page 8.
There is enough information there to work this out.

Page 8? Looks like a flower petal with numbers around it.

I thought Kevin had decided to go with a 12 V DC supply, in which case none of that matters.

Redbelly- Thanks for keeping that clear, I was starting to wonder if that was all just to run on DC or not.


Like I said earlier, I knew this was the place to figure out this little wiring question. I just didn't know it would spark such a "discussion"

I now know that I can string 200 LEDs together and I know that I can run them on a 12 VDC power supply, which will make building it soo much easier.

 

[Redbelly98]

Page 8? Looks like a flower petal with numbers around it.

It's the page that says "8" on it, which is p. 9 in the .pdf file numbering.

 

[Phrak]

Try looking at Page 8.
There is enough information there to work this out.

If you can't tell me where you got a count of 44 from, then I don't see any point to continuing this conversation.

I'm sorry to have frustrated you NoTime. I think it has become rather pointless, but from page 8, I assume I can't rely on the Duty Ratio plot, and use the max rated current of 30 mA instead.

\frac{(1.414)120V - 1.4V}{44} = 3.82V

At 3.8V per diode the forward current is about 30 mA.

 

[NoTime]

Mind you, I have nothing against side-discussions and I think they're partly why Physics Forums is such a fantastic place. But I did want to clarify that for Kevin's sake.

Side discussions are fine by me.
Originally, I just wanted to clear up some ambiguities with Phrak's schematic for the benefit of other people that might be reading this.
LED's are a fairly popular topic.

Good point about the clarification.
It's a good thing you mentioned this

 

[NoTime]

Page 8? Looks like a flower petal with numbers around it.

The flower petal is a color map.
Here is one painted in with the colors.
http://www.google.com/imgres?imgurl...n&sa=X&oi=image_result&resnum=3&ct=image&cd=1

It's the next page, I was referring to.

Redbelly- Thanks for keeping that clear, I was starting to wonder if that was all just to run on DC or not.

Only the part about dynamic resistance is important to you.
You can use 12v, 3.6v and 30ma in the calculator link.
Your project will work, but the final measured current will be around 10ma rather than the expected 30ma.

Have fun

 

[NoTime]

I'm sorry to have frustrated you NoTime. I think it has become rather pointless, but from page 8, I assume I can't rely on the Duty Ratio plot, and use the max rated current of 30 mA instead.

\frac{(1.414)120V - 1.4V}{44} = 3.82V

At 3.8V per diode the forward current is about 30 mA.

That explains a lot!
Nice touch with the bridge diode drop, but I will note that the nominal line voltage, in the US, is 118v rather than 120v. It also varies some.

I don't think this side discussion is pointless.
Since you showed your work it isn't frustrating either

Quiz questions:
Compute the RMS current for the 100ma 10% duty cycle square wave given in the data sheet.

Compute the duty cycle for your solution (120hz).
Assume a cutoff voltage of (where no current flows) of 3.819v.

Compute the duty cycle of the portion of a 60hz sine wave that exceeds the 120v RMS value.
Assume only the positive half of the full cycle conducts (no bridge or switch diode).

 

[NoTime]

You can use 12v, 3.6v and 30ma in the calculator link.
Your project will work, but the final measured current will be around 10ma rather than the expected 30ma.

I made a mistake here.
Use 20ma in the calculator link.

This will work out quite closely to 20ma (about 17ma) using the 68 ohm resistor the calculator will show.
The reason it works is because the 3.6v is Vf at 20ma.
The calculator shows standard resistor values instead of exact values, or why the final value is not exactly 20ma.


My mistake was to use the 3.6v (Vf @ 20ma) instead of V_threshold which is about 2.9v.
If you use the 47 ohm resistor then the final current will be around 25ma, somewhat distinct from 10ma.

 

[Phrak]

Talk to me as if I have 23 odd years in electrical circuit development or design in power conversion, instrumentation, controls, fixturing, and fixing other peoples design errors, neither pushing papers nor peripheral support, interspersed with firmware coding and software engineering in test, engineering tools, and simulators, with exerience in primary and root cause failure analysis.

 

[NoTime]

Talk to me as if I have 23 odd years in electrical circuit development or design in power conversion, instrumentation, controls, fixturing, and fixing other peoples design errors, neither pushing papers nor peripheral support, interspersed with firmware coding and software engineering in test, engineering tools, and simulators, with exerience in primary and root cause failure analysis.

Fair enough.
Part of what I said to you made me realize what I had forgotten.
So I give you credit for fixing my mistake.

Your number of 44 also solves for a 20ma RMS current, except you got the count right.

One of the good things about PF is that if you do get things wrong there are people to set you on the right path,

Thanks.

 

[Phrak]

Fair enough.
Part of what I said to you made me realize what I had forgotten.
So I give you credit for fixing my mistake.

Your number of 44 also solves for a 20ma RMS current, except you got the count right.

One of the good things about PF is that if you do get things wrong there are people to set you on the right path,

Thanks.

You're a better man than me, NoTime.

I've been too under the weather to respond lately. I may still be too dopey.

We don't really see this exexcise as the same problem. In my view, the objective is to obtain the most luminosity with the least component count and cost, without thermally degrading the components. Thermal degregation is everything. This means keeping the peak current within bounds while delivering the highest possible average current. These things are diodes, and these sort of curcits the dynamic impedence is not a useful number given the wide swing in bias voltage. The current through one of these LEDS varies roughly as the exponent of the bias voltage, I'm sure you know. Roughly,

I = exp(kE)-1

P = c(E exp(kE) - 1)

So as the voltage increases from zero the current creeps slowly up with a ugly shape peak in as the AC peaks. The idea in any design is to keep the peak instantanious power down. But it's a little more than that. Thermal degragation is usually about wire bond integrity and more importantly mirgration of dopants at the junction. In this case you might add degregation of the fluorescent dyes. Beats me. The curves you see in the data sheet only hint at the thermal limits. You might ask yourself why these boundries in their Allowable Forward Current area plots are drawn where they are. There are six different boundries going on; I don't know where they all come from.

You know about thermal circuits right? In any case this is all for general consuption anyway. The heat driven into the die is the instantanious resistive power lost integrated over time. The heat conducted out depends on the bulk thermal resistance as the heat is conducted away via leads and plastic case. So in the simple situation of single pulse power, for instance, the instantanious temperature of the die ramps up like voltage across a capacitor in parallel with a resistor, being charged with a current source. To make a long story short, after approximating the repetative current duty cycle if you know the normalized thermal frequency response curve you obtain the heating obtained as if it were heated by a constant DC bias.

As there is no frequency response curve provided by the manufacturer, we can only guess based upon where the boundries show up in the Allowable Forward Current plots. The thermal response curves are based upon duty cycle and frequency. As Nichea Corp was not gracious enough to let us know the test frequency used in obtaining their Duty Ratio vs. Allowable Forward Current plot, we can just guess or defer to experimentation. [Edit: If it's 10KHz or better which is pretty typical for strobed character/segment multiplexing, then it's useless. If it's 1000 Hz, which should be more typical for scrolling displays that use descete diode assemblies it could be somewhat usefull.] The latter (experimentation) is what I recommend. The 44 diode count solution I had is pretty dismal: assuredly within max allowable current spec, but probably about as bright as running on 10mA DC bias.