How do I exend supply range of LM317 for CC mode?

[throbscottle]

Hello good people
I've been building UV LED light box for making PCB's. Going well so far. Based on a project I saw on Instructables, it will have 198 LED's split into 3 sets of 66 in series. Each set has an LM317 set to deliver 18mA as a constant current source. Each LM317 has a resistor in series with its input to drop a chunk of voltage. It runs directly off rectified/smoothed mains voltage. I based the design on a nominal mains supply of 240V. Each LED has nominal Vf of 3.3v, so it's about 218v per section.
I tried it with 1 string of 66 LED's, and that section works really well. But now I've hit a snag...
The tolerance for UK mains voltage, I checked today, is +10% and -6%. So the DC going into the thing can be 306v to 358v, a span of 52v!
So I've been trying to see how I might lose 15 to 20v off that range without building an actual regulator or resorting to a high voltage zener, nothing leaps out at me, so I thought the most straightforward thing would be to extend the supply range of the LM317's somehow.
Any thoughts? Anyone?

 

[throbscottle]

Actually I think I have found a partial answer - a floating LM317 regulator with a zener between it's input and output, like this: https://electronics.stackexchange.c...at-high-voltage-protected-by-two-zener-diodes except I'd be using a 36v zener. I just need it to turn into a short circuit when the mains drops too low...
Or conversely, set the regulator to the low end, and let the zener take some of the current when the voltage is too high. Any suggestions how I find how much current it would actually be carrying? Given a total load of around 54mA. Brain hurts.

 

[crutschow]

How about an **broken link removed**, a high voltage (100V), low current (50mA) regulator with the same configuration as the LM317?

 

[Tony Stewart]

FYI

Input = 358Vdc max 306Vmin
Load = 218V @ 18mA
Difference = 88V min to 140V max

You could easily add a few K , 2W between diode bridge and filter cap to reduce the Vdc min to 3V .

Or use FETS rated for 200V with this classic CC limiter

Why not add more LEDs? no need?

 

[throbscottle]

I'd never even thought about high voltage regulators. 1 track mind.
The reason for the strange number of LED's is that the design I've based this on has 200 LED's and apparently gives very good performance. It has a separate dropper resistor for each LED and runs on 12v - lots of wasted power. So I went for constant current, series connection. So I had a choice of doing 2 higher voltage strings and 1 lower voltage one, various much lower voltage ones, or 3 moderately high voltage ones, the option which looked the best, and would be if the darned mains voltage wasn't quite so variable.
Anyway I got 10 LM317's cheaper than getting the 3 I actually need.
So if I'd thought more about it Tony your regulator would be just the ticket. I'm using this circuit for something else and really should have thought of it. Definitely the one to use if starting from scratch again, but as it is I'm reluctant to take out what I've already got.
OTOH, the LR12 is a little hard to obtain economically, but I see there are some others to be had, eg TL783 and ... oh that's about it as cheap goes...
(OK been to work and come home since starting this reply! Also had a bit of a goooooogle in-between the work bits.) I think this is what I need to use if I want to keep the existing circuit so far since I can build it with parts I already have (just not the same ones as in the AN and set it for 222v) : **broken link removed**
drat that means I'll have 3x 4k7 2W resistors I'm no longer using...

 

[throbscottle]

**broken link removed**

Been having a play with it in LTSpice. Found it to be quite sensitive to the values of resistors in the pd. R2 at 560k and R1 at 10k with a 3v3 zener gives me 240vdc. Can't seem to get a voltage between 220 and 240v

 

[AnalogKid]

I use a variation of that circuit for reverse polarity protection and input overvoltage clipping in a MIL power supply. In your case the FET sill not be fully enhanced (saturated) so it will get hot 20 mA at 140 V = melted fingerprints.

ak

 

[Tony Stewart]

Perhaps a simple buck regulator current source is your best bet if you also want it cool.

 

[throbscottle]

Looking at the current through the mosfet it's either on or off - and mostly off in this case. Carries about 1.3A when on. Cap is 47u, not the 4.7u shown in the AN Total load current will be about 54mA (3x 18mA).
Not too bothered about heat so long as it's not excessive - it won't run more than a few minutes at a time.
Again, buck regulator - something I could/should have done starting from scratch, but I have what I have.

 

[MikeMl]

Back to what you posted as requirements in post #1: The voltage standoff requirements for the current sink are more demanding than you thought. Here are 66 Leds with a nominal Vf=3.3V@18mA. I vary both the temperature (TEMP) and the source voltage 306V<V1<358V.

The upper plot pane shows the voltage drop across the current sink V(sink) at three different temperatures Green=0 degC, Blue=25 degC, Red=50 degC as a function of V1.

The Lower plot pane shows the temperature variation in Vf of a single diode @18mA.



The power dissipation in the current sink is shown below, so it will have to be adequately heatsinked...

 

[spec]

FYI

Input = 358Vdc max 306Vmin
Load = 218V @ 18mA
Difference = 88V min to 140V max

You could easily add a few K , 2W between diode bridge and filter cap to reduce the Vdc min to 3V .

Or use FETS rated for 200V with this classic CC limiter

Why not add more LEDs? no need?

Tony this circuit has a number of problems.
(1) It is overly complicated: just one NMOSFET will do the job
(2) The constant current is ill defined: the gate threshold voltage of an IRF640 is 2 to 4V and that is ignoring temperature effects, so the constant current will have spread of 2:1
(3) R4 is incorrect: should be 170R. 180R is the closest standard resistor value. This is to give a constant current of 18mA with the average IFR gate threshold of 3V.
(4) That combination of PMOSFETs is certain to oscillate: there are no gate stoppers.
(5) The characteristics of the reservoir capacitor are not specified.
(6) There is no LF or HF decoupling
(7) The circuit is a safety hazard on a number of counts but at least it needs a mains fuse.
(8) The working voltage of R1 is not specified and will exceed the voltage rating of common resistors

spec

 

[schmitt trigger]

........... It has a separate dropper resistor for each LED and runs on 12v - lots of wasted power................

I see this design flaw repeated in many "commercial" products.
They tout the LED's significant efficiency advantage over an incandescent.
Only to waste it all on dropping resistors.

 

[throbscottle]

The upper plot pane shows the voltage drop across the current sink V(sink) at three different temperatures Green=0 degC, Blue=25 degC, Red=50 degC as a function of V1.

That's quite an eye opener, Mike. Luckily I'll be using it round about room temperature and it won't have chance to get particularly hot. Looks like I really need that supply voltage limiter though...
Going back to the original design, each current source had a 4k7 2W resistor in series with it to drop most of the voltage.

 

[Tony Stewart]

I agree the TL783 (125V) is the simplest solution, with suitable voltage drop R and ( don't forget 500V rating) from 352V bridge peak to cap. with suitable holding current for 10-20 ms.

warning, Only for experienced hobbyists, working with rectified line voltage and double insulation and use of Polyfuse protection.

 

[spec]

I agree the TL783 (125V) is the simplest solution, with suitable voltage drop R and ( don't forget 500V rating) from 352V bridge peak to cap. with suitable holding current for 10-20 ms.

warning, Only for experienced hobbyists, working with rectified line voltage and double insulation and use of Polyfuse protection.

While this is a good approach, even the TL783 will not do the job because it only has a maximum input output differential of 125V and around 162V or possibly more is required by the constant current generator.

The only minimum solution I could come up with for a high-voltage constant-current generator is similar to Tony's post #4, except with a bipolar sense and pass transistors and a high side, rather than low side configuration as shown below.

In effect, you would be constructing your own LM317 as shown on the schematic so the home made LM317 could simply replace the LM317 in your original circuit. The current sense resistor is calculated by IK= 0.6V/R-3 and needs to be the value shown in the schematic for 18mA Ik.

spec

WARNING: THIS CIRCUIT HAS HIGH VOLTAGES AND SHOULD ONLY BE CONTEMPLATED BY THOSE EXPERIENCED IN HIGH VOLTAGE ELECTRONICS. ON NO ACCOUNT TOUCH ANY PART OF THE CIRCUIT WHILE THE CIRCUIT IS PLUGGED INTO THE MAINS SUPPLY

​

NOTES
(1) C1 and C2 are 400V minimum disk ceramic (not surface mount) capacitors X7R dialectic +-10% or better
(2) R-1 is 400V minimum voltage.
(3) The pass transistor Q-1 needs to have a VCEo of 400V ideally and be capable of dissipating 2.9W which, without a heatsink, implies a TO220 type. There are many transistors that would meet these requirements.

 

[MikeMl]

Perhaps Tony was trying to post something like this:

 

[throbscottle]

I don't see the problem with using the TL783 with dropper resistor in series with it? That way it only has to handle the input voltage span + variation due to temperature of the LED's, ie 66x3.0=198 say is worst case for heat of LED's, 358v worst case for mains, gives me 160v. The LED's are actually sold as having a Vf of 3.5v, so using that gives 66x3.5=231 as the cold value, mains low voltage gives 305, difference is 74. So say I can lose 70v across a dropper resistor, that leaves 4v for the TL783, and in the worst case it's then 160v-70v, gives 90v.
Not very professional but then, I'm not a professional
I still think I'll try that limiter circuit before I change these regulators though. It's inspired me to have just spent a happy hour sorting out my collection of old board pulled zeners so I actually know what I've got!

 

[AnalogKid]

Another approach is a non-standard version of a hysteretic regulator. Fullwave rectify the incoming AC but *don't* filter it, and have a pass device that is turned off when the incoming sine exceeds 218 V or whatever. Harris had an offline 5 V regulator based on this approach back in the 80's. The LEDs get four current pulses per power line cycle, more than enough to prevent perceiving flicker, and there is almost zero heat in the pass device. An extension of this method has the circuit ahead of a filter cap, followed by a current limit resistor and the LEDs.

ak

 

[throbscottle]

Umm, AK, that's what the limiter does...

 

[Tony Stewart]

recall
Difference = 88V min to 140V max
Keep in mind the Vf and thus ESR of the UV LEDs will have a wide tolerance, (E.g. 25%) unless yours is already measured and you plan on only making 1.

A series 1/2W 500V resistor can be selected for the bridge filter and effective load to drop the Vmin by 80V (e.g. 80V/20mA=4K max. choose nearest lower value.

Then any CC design capable of 8-60V is needed.
The series R will have ~350Vp Max on startup and thus dissipate a 32W pulse and should be wire-wound ceramic, non-flammable.

The other Designs must consider the surge current in the storage cap and diodes for pulse current ratings on startup. This one is current limited to under 100mA.

I just read @throb... had the same plan. I agree... and I am now a retired professional since starting in 1975.