LED drivers with very high analog dim range?

g

Your specs keep evolving as you add constraints, making the target a little elusive. But so far, if you need 4A, high efficiency, and

1000:1 range, you could

1) Sum an LM3409 and a linear analog current source into the LED or

2) Make a SMPS voltage source, set Vout =3D Vf(led_4A) + XXXmV, then control current with a low-side analog current sink. You could even modulate Vout with Iout a little, to improve efficiency.

3) Sum the output of two LM3409 ckts, one made for high and for low current, DAC-style.

#2 seems easiest. It lets you use std voltage-mode SMPS chips, so it's dirt cheap.

-- Cheers, James Arthur

Analog

"digital

marketing

cheaper:

0.5%

deal

easy.

then?

True. But since the chip is limited to 10:1 even for PWM dimming I don't see any motivation for them to include offset trim inside that circuitry. Which would be needed for very large analog dimming ranges. Offset trim is an extra production step and thus cost in IC manufacturing.

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

You need to use clean switch-mode in this business because the "my light has 0.5% more efficiency than yours" battles are rather fierce.

Also, the LED industry is notoriously making the mistake that they compare their lights to incandescents, claiming 80% power savings or whatever. The reality is this though: They must compete against CFL. Today a 60W-comparable CFL runs at 13-14W. A 60W-comparable LED costs several times as much and uses 12W. That just ain't good enough.

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

Yeah, sorry, but I didn't want to turn the first post into a whole sermon :-)

Essentially I'd be also happy if someone know of any LED driver IC that offers more than 250:1 analog dimming range. Regardless of efficiency.

Regular converter ICs and discrete stuff is what has been used so far (at least I did). Works, but it's lots of parts. I was curious as to whether there is a real LED driver chip that has all this in there and can do more than 250:1 dimming range natively, not with a lot of helper circuitry.

--
Regards, Joerg

http://www.analogconsultants.com/

log

l

ng

d

There's not going to be a 120VAC-to-4A converter with a 1000:1 linear dimming range. Nobody needs that.

A standard SimpleSwitcher type supply is awfully simple. Adding an op amp, FET, and sense resistor hardly seems like a lot of helper circuitry. It's gonna be hard to get simpler or cheaper than that.

-- Cheers, James Arthur

nalog

tal

ting

I though you said you wanted analog, not sm. Maybe a misunderstanding.

NT

Oops, sorry, "analog dimming" is the LED industry expression for dimming via smooth current regulation, modifying the duty cycle in the main converter. In contrast to PWM dimming where the whole converter is turned on-off at a lowish frequency (usually a few hundred Hertz).

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

Well, some market do need that. But not in homes, of course. And it's ok if the IC can't support PFC or whatever, all I'd like to know is if there is any IC better than that National (or now TI) LM3409.

Not as easy as you think. Let's say you drop 100mV at full power. Now if you'd regulate down to 1/1000th or 0.1% you'd be looking at 100uV. To hold that with any amount of stability in an environment with the usual noises, cell phones and whatnot, you need to do this differentially and probably with a CMOS amp. Needless to say, it also requires a chopper amp because no reasonable opamp will afford that low of an offset. That leads us into the loop stability issues that chopper amps can bring about ... it's opening a can of worms :-)

--
Regards, Joerg

http://www.analogconsultants.com/

I agree.

The perceived brightness is approximately square root of intensity. The linear range of 1000 is only 30 times of visible difference. If the seamingly smooth adjustment is required, the number of dimming steps should be large, probably in the hundreds range. Those requirements are not very easy to implement with PWM.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

nalog

tal

ting

t

and

n r p

I believe I understand what the application is now. If it's what it seems to be I've solved this same problem already, cheaply, but it's proprietary to me, and a conflict. I'll have to recuse myself...

-- Cheers, James Arthur

Understandable :-)

--
Regards, Joerg

http://www.analogconsultants.com/

Analog

fairly

"digital

marketing

that

efficiency.

and

helper

You may want to look at PFC chips themselves. I expect that some PFC chips are used to provide dimmable led lamps on the market today.

?-)

[...]

However, then I am back to doing all this feedback stuff inlcuding a good reference and all that myself, externally. That's what I wanted to avoid.

--
Regards, Joerg

http://www.analogconsultants.com/

te:

op

if

l d t

What about a microcontroller. Perhaps the Cypress PSoC 1. Internal reference and opamp. Crappy op-amp that is, but if you do coherent double sampling, combined with production calibration, you can do everything inside that IC, with just the power components externally (FET, inductor, diode and cap)

A lot of standard micro's with analog frontend can do this. The only catch is that you need to be able to sleep nights, knowing you have cheated, that is trusted a production calibration, and you didn't do any "real" analog stuff

Regards

Klaus

That is a good idea. I've thought about dual-slope schemes and such as well but those are slow. The PSoC could do this faster. A risk is the fact that the opamps are on a chip that, in contrast to a switcher, runs very irregular digital stuff. So you'd have to cross your fingers that the noise during the second sampling is at least very similar to that of the first. To some extent this can be alleviated by programming "close to the machine", maybe in assembler or at least inline assembler. Maybe also by having a second "dummy opamp" track the on-chip noise and compensate it out to some extent.

But the real challenge is the rather horrid voltage noise in these opamps:

That would make reliable measurements and tracking at 100uV or less a real challenge at the usual loop bandwidth of a few ten kHz.

And hope that they don't switch foundries or processes, the noise gets a much different characteristic, and then the phone rings ... :-)

--
Regards, Joerg

http://www.analogconsultants.com/

rote:

an op

t.

w if

o

ual

and

r

hat

FC

o :

For a fixed setpoint, the offsets of the PGA/ADC won't change much over time, so you could probably get away with only doing the zero calibration sample once in a while, to boost the bandwidth

Question: Why do you need fast bandwidth, for a human interface thingee, LED lamp?? Is it due to input voltage changes? If so you can do feedforward, but if its current mode control, then you don't need feedforward....

You could use a DAC (also in the PSoC), to do a simple sigma delta feeding the voltage from the DAC and the sense input into a comparator. Then you regulation is done via a comparator which is fast. You then need a slower outer loop for DC regulation. The microcontroller advantage is that you can sweep the different parameters during production and save a look-up table. For a load shift, you can do aggressive PID regulation, since you know pretty much where you are and where you need to be and the characteristics of the SMPS, so predictive regulation can be done (almost better than a critically damped system)

The noise looks bad. Perhaps you can do some averaging in the outer loop?

Regards

Klaus

te:

op

if

l d t

how will you be controlling it, must it be an analog voltage or can it be a digital interface?

-Lasse

For many applications a slow loop would do. But there are some situations where rapid dim up/down is required and that needs something in the low tens of kHz.

I had suggested a dual loop approach, with the outer loop sufficiently slow so that a dual slope conversion can work which is very cheap. Mainly because I am already familiar with double and even triple loop designs from PLLs. But it was not met with enthusiasm, it is probably too unorthodox or maybe it has too much of a "duct tape feel" to it :-)

Partial forward control as an alternative was also not exactly met with standing ovations.

That might be tough with a PSoC. The very low frequency stuff looks more like continuous little earthquakes going on in there. I think a dual-slope approach for the outer loop could work, just not sure if an external comparator would be needed to reach well above 60dB.

--
Regards, Joerg

http://www.analogconsultants.com/

g an op

hat.

Now if

To

usual

y and

per

That

g

PFC

to

ns

of

e

mps:

a

So they want to have it all for no bucks, but no tricks ;-)

On the subject of the comparator. Could a digital input (not schmitt trigger) be used as a fast comparator, ofcourse using the VDAC for the offset? Outer loop still slow DC....

Regards

Klaus