Dimmable LEDs

Sorry if this is here already, didn't find it.

What makes a GU10 or similar LED lamp "Dimmable"? Given that a simple LED works on 3volts or so there must be something in the enclosure to allow 12V working. What do they use?
Switching regulator?
Resistor?

Cut one open?

I need a bright dimmable COB style, preferably about 10W. The type with a lot of separate leds won't suit.
(One with a good CRI would be best)

ta

Mains-voltage dimable LEDs assume that they are being dimmed by a traditional triac dimmer and have a circuit that observes the input waveform to determine when the triac turns on and hence the dimmer setting. They then have a switch mode supply, and adjust its output to the required brightness.

12V LEDs either have several LEDs in series (e.g. Led strips) or they have a switch mode supply for a single led. The individual LEDs in a COB led module are in parallel IIRC, so they would need a switch mode supply. I hope no-one uses a resistor!

I'm not entirely sure what dimable 12v LEDs do. If they incorporate a switch mode supply they probably try to determine the required brightness by observing the input, like the dimable mains LEDs.

If they just used a trad triac they'd flach at 50Hz? Useless for what I need.
There are mains "LED dimmers" aren't there? They would presumably either PWM chop at a high frequency, or reduce the current and stay on.
Then every LED would be dimmable. SO wouldn't they just say that?

To use a led with a trad dimmer would need a big capacitor to fill in the no-volt time, surely?

So I'm still not clear what 's in the "lamp", (except that I'm aware there were some with just a resistor, which got hot and went o/c).

ChrisR said:

If they just used a trad triac they'd flach at 50Hz?

Click to expand...

Yes if they "just" used a traditional triac dimmer, but as I wrote they have a circuit that observes the waveform from the triac dimmer and determines what it is set to, and then adjusts the switch mode output to get the required brightness.

Here's an example:

http://www.nxp.com/documents/user_manual/UM10394.pdf
Section 4.2 describes the triac phase detection circuit.

What exactly is your question? Do you want to understand how they work out of curiosity?

Thanks for the reference.
I'm curious, yes, because I need a bright LED which is dimmable and doesn't switch off for a chunk of the 100Hz cycle. Those dimmers don't show a waveform, but I'm assuming that they work by controlling the current during the "on" part, of the cycle which may be say 20%, so for 80% of the period they're off.
This isn't an area of electronics where I'm comfortable - it may be that there's enough flyback energy to fill in the "off" period but I don't think so, it would still need a big capacitor.
SO the next question is how do "proper" mains led dimmers work, and if it's by reducing current then why don't they dim any LED? If Non Dimmable LEDs could be used with them , I think it would say do on the LED box.
See what I mean?

The LED is for an epi-illuminator in a microscope. I really would like a PWM at >100kHz.
The thing weighs 100kg+ but still vibrates enough at high magnification that I want to use as short exposures as I can. At 1000th second that's only 10 pulses, so subsequent exposures would vary by a few % depending on phase.
I'm using several hundred exposures per image and don't want the intensity changing much.

Yer tiz. Lamp will be in the crude looking black box on the left. Camera sits on top where this one shows a video cam with a yellow label.
Stepper drive on the focus knob (black, rhs, not fitted here) gives 1 micron vertical steps. Encoders in base give 1 micron x-y position.
40 odd kg of granite slab also not shown here.

The input capacitor on a AC-input switch mode supply will definitely be large enough to cope with the off period caused by a triac dimmer. But that apparently doesn't necessarily mean that there won't be 100 Hz flicker, see below.

Here's an interesting application node, about implementing a 12V MR16 LED lamp that works with existing "electronic transformers" and dimmers:

http://www.maximintegrated.com/en/app-notes/index.mvp/id/5372

Have a look at figures 7 and 8. They show the current through the LED when this control circuit is used with a dimmer. As you can see, they copy the waveform created by the dimmer so the LEDs will flicker at 100 Hz. This is not how the NXP design in my previous post works!

I don't know what LED lamps that you can buy actually do. It's quite possible that they are much cruder than either of these chip manufacturers' reference designs.

If I were you, I'd try to get a raw LED module, fit a suitable heatsink (important!), and then connect it to a lab power supply or similar with an adjustable constant-current output. That way you know exactly what is going on.

Good luck!

endecotp said:

If I were you, I'd try to get a raw LED module, fit a suitable heatsink (important!), and then connect it to a lab power supply or similar with an adjustable constant-current output. That way you know exactly what is going on.

Click to expand...

That is the only way you will get smooth , flicker free and controlled dimming from a LED light source.

Nearly every good 230 volt LED will turn off and on as nearly all good LED's use some form of switch mode control. The word switch in switched mode is the clue rather than vary the resistance which would produce heat they switch the supply on and off. Since the LED is DC it could have a capacitor to smooth the output but in real terms you have not got a clue as to exactly how each LED lamp is regulated.

As you say a white LED is around 3 volt threshold voltage with a red one around 1.2 volt but they are not controlled by voltage it is the current which is controlled. For a LED warning lamp often a simple resistor is used but these will produce heat so since whole idea of LED lighting is not to produce heat one hopes they do not use a simple resistor. However at 12 volt using three LED's in series and one resistor one can make a cheap unit. With pineapple style even with 230 volt using 70 LED's and a resistor is an option and it would seem this is done with some makes.

However better makes use switch mode regulation and as a result common for input voltage to be 150 ~ 250 volt. So the lamp monitors the wave form and uses this as a signal there is then a whole range of options which likely include turning into DC what it would seem is important to you is if it has a capacitor included but short of reverse engineering the lamp you don't know if the lamp actually flashes at some high frequency or not. I suppose you could use some light sensitive device connected to an oscilloscope, but if you need to know it is smoothed DC driving the LED then really only way is home built drivers.

The other thing about LED light is that the spectrum of a "white" LED ( white to the human eye ) may not have the full spectrum that a camera needs. I don't fully understand why but it has something to do with the difference between colour by substraction / absorbtion and colour by addition of different light wavelengths

Yes a SMPS switches but continuously at some high frequency. A dimming one can work by altering the mark/space ratrio of the waveform, and/or by controlling the current during the On time. But a phase controlled triac dimmer works by turning off for some large fraction of the 100Hz cycle, like half of it. That's the "off" bit I can't tolerate.
If the "switching" of the SMPS is at >100kHz continuously than it's not a problem because of integration over the exposure period.
Yes, I've made things like this


but mounting them is a bit of a pain compared with using an off-the-shelf GU10 "bulb" in an anglepoise - or something.
The one in the pic is a nice bright one with a CRI (see first post) of about 97 (That's the Spectrum bit)

"The input capacitor on a AC-input switch mode supply will definitely be large enough to cope with the off period caused by a triac dimmer."
Not sure what you mean. If it's off, it's off.

--

Hence wondering what's in the dimmable vs non dimmable "bulbs" - apart from the crappy Chinese ones with resistors!
I'm thinking that the non-dimmable ones have horrible negative reistance characteristic or something, which sets up an oscillation which leads to visible fllickering, but I'm not sure. I'll take one apart some time and see what's in there.

I can't use the very wide led arrays because of the size of the parts.
(cf Köhler illumination)

ChrisR said:

"The input capacitor on a AC-input switch mode supply will definitely be large enough to cope with the off period caused by a triac dimmer."
Not sure what you mean. If it's off, it's off.

Click to expand...

What I mean is, the input capacitor (after the bridge rectifier) in an AC-input switch mode supply is normally quite large. So a design like the NXP one in my first link above where the LED current is constant throughout the 50Hz cycle works OK, because the capacitor stores enough energy to last through the part of the cycle where the triac is off.

Hence wondering what's in the dimmable vs non dimmable "bulbs"

Click to expand...

Well that's what I've been explaining....
They all have a switch-mode supply. For the non-dimable ones, that's all they have. For the dimable ones, they have some additional circuitry that adjusts the output based on how the input has been dimmed. The two designs that I've linked to show two different ways to do that.

I'm thinking that the non-dimmable ones have horrible negative reistance characteristic or something, which sets up an oscillation which leads to visible fllickering

Click to expand...

No. The reason why non-dimable lamps flicker (at maybe 1Hz) when connected to a dimmer is that they keep shutting down and then starting up again because the input voltage is too low. All switch mode supplies need some sort of bootstrap power supply to start the oscillator. The period of the flickering is a result of the time for the bootstrapping and oscillator startup

endecotp said:

the capacitor stores enough energy to last through the part of the cycle where the triac is off.

Click to expand...

OK, what sort of value cap is that going to be though? Depending what ripple is tolerable, I get something of the order of 10,000uF. Seems a bit large to go in a thing like:
http://www.miniinthebox.com/gu10-10w-5-2w-3000k-warm-white-dimmable-high-power-led-bulb_p791974.html ?

0/10, no working shown

For a 10W lamp you would need to store 0.1J to last for 1/100s. Rectified mains is about 340V. Energy stored in a capacitor is E = 1/2 C V^2 so C = 2 E / V^2 = 2 x 0.1 / (340x340) = 1.7 uF.

For a 12V system it needs to be larger, I get 1400 uF. But the capacitor only needs to be rated for maybe 20V so it's not a thousand times larger than the 1.7 uF 400V part.

In reality you'd use something larger than that.

Maybe this is a reason for the different approaches in the two designs I linked to, as the second design doesn't need a large input cap; it just needs to store enough energy to keep the oscillator running.

1v at 1A ripple gives
C=I*t/ΔV
=1a *10ms/1V
=10000uF.

I think you're suggesting the cap can be completely discharged at the end of the 10ms. What's that going to do to the regulation?! I need a stable HF output continuous output.
But/and - though the more clever regulators may keep the oscillator running - both the cited ones seem to do that - the light still goes out, = no good to me.
Upshot is that though there are interesting things in an LED "bulb", I can't use them!
I'm now mre curious about how they dissipate the heat. I may cannibalise one to remove the electronics and leave the ironmongery.

There ARE people with microscopes using a hairy great rheostat to dim their LED, once they've got the voltage down close. Not so daft really, it keeps things simple!

I'm not sure what ripple you're referring to, nor where 1V or 1A come from. Changes in the input cap voltage won't affect the LED brightness as the switch mode control loop should be fast enough to adapt to it.

The cap can be mostly discharged each mains cycle, if you're careful about the rest of the design; I'd allow a factor of at least two to be safe.

The first cited design (the mains input one) maintains a constant current through the LED, it does not go out.

They disipate heat with a big lump of aluminium and then make sure the control circuitry will function at 80-100 C.