What are some good switching-mode LED drivers, able
to work > 26V, regulate current at 30A (ext FETs),
and do fast PWM for 3000:1 or better dimming? We
looked at MAX16819, what are some others you like?
--
Thanks,
- Win
LED drivers, >24V, 20-30A current reg, fast PWM
Started by ●April 3, 2019
Reply by ●April 3, 20192019-04-03
On Wednesday, April 3, 2019 at 1:55:44 PM UTC-4, Winfield Hill wrote:> What are some good switching-mode LED drivers, able > to work > 26V, regulate current at 30A (ext FETs), > and do fast PWM for 3000:1 or better dimming? We > looked at MAX16819, what are some others you like? > > > -- > Thanks, > - WinWhat kind of PWM circuit are you contemplating for 3000:1?
Reply by ●April 3, 20192019-04-03
bloggs.fredbloggs.fred@gmail.com wrote...> > On Wednesday, April 3, 2019, Winfield Hill wrote: >> >> What are some good switching-mode LED drivers, able >> to work > 26V, regulate current at 30A (ext FETs), >> and do fast PWM for 3000:1 or better dimming? We >> looked at MAX16819, what are some others you like? > > What kind of PWM circuit are you contemplating > for 3000:1?In addition to their current-mode control, the chips have a PWM input pin, and claim 2000 to 5000:1 dimming capability. In the fine print you discover you have to slow the PWM rate significantly, and also deal with a minimum pulse length, plus uncertainty in the on/off delay times. Consider, with a 20kHz PWM rate, a 100ns width uncertainty limits you to a 500:1 range. But if the PWM rate was 3kHz, you could get 3000:1 range. To get zero, make no pulses, or to get a light level of 1 part in 3000, make a single 100ns pulse, every 330 us. I'm not very concerned about that spec. -- Thanks, - Win
Reply by ●April 5, 20192019-04-05
Winfield Hill wrote...> > What are some good switching-mode LED drivers, able > to work > 26V, regulate current at 30A (ext FETs), > and do fast PWM for 3000:1 or better dimming? We > looked at MAX16819, what are some others you like?Luminous PhlatLight uses a MAX16818 in their 36-amp DK-136M evaluation drivers, for their big projection LEDs. We used that with good success. The MAX16818 looks good, $3.45 each, except it's already NRND, and Maxim's recommended replacement, the MAX20078, is NDA-required, just to see a datasheet! And it's an especially onerous NDA, you need to be a very high-volumn user, and an officer of the corporation with sufficient authority, has to sign the NDA, etc. I'm still searching for some candidate parts, so far there's only the LT3743. This also worked well for us, but it's a bit of a complicated design. -- Thanks, - Win
Reply by ●April 5, 20192019-04-05
Roll your own with a microcontroller and your own gatedriver If a lot more work off course , but fun Cheers Klaus
Reply by ●April 7, 20192019-04-07
On Wednesday, April 3, 2019 at 7:48:41 PM UTC-4, Winfield Hill wrote:> bloggs.fredbloggs.fred@gmail.com wrote... > > > > On Wednesday, April 3, 2019, Winfield Hill wrote: > >> > >> What are some good switching-mode LED drivers, able > >> to work > 26V, regulate current at 30A (ext FETs), > >> and do fast PWM for 3000:1 or better dimming? We > >> looked at MAX16819, what are some others you like? > > > > What kind of PWM circuit are you contemplating > > for 3000:1? > > In addition to their current-mode control, the chips > have a PWM input pin, and claim 2000 to 5000:1 dimming > capability. In the fine print you discover you have > to slow the PWM rate significantly, and also deal with > a minimum pulse length, plus uncertainty in the on/off > delay times. Consider, with a 20kHz PWM rate, a 100ns > width uncertainty limits you to a 500:1 range. But if > the PWM rate was 3kHz, you could get 3000:1 range. > To get zero, make no pulses, or to get a light level of > 1 part in 3000, make a single 100ns pulse, every 330 us. > > I'm not very concerned about that spec. > > > -- > Thanks, > - WinIf you're going for 3000:1 and you want an accuracy of +/- 1/2 least significant increment, which would be 1 part in 6000, then an uncertainty of 200ns in the rise/fall would make for about 400ns minimum programmed pulse width. So that 3000x would be 1.2ms= 833Hz clock. That gets you well beyond any problems with flicker. The only problem is the illumination engineers define dimming differently from the semi people. The higher power LEDs are relatively sensitive to wandering around their color coordinate space as a function of drive current, and supposedly human vision is just as sensitive to picking this up. So when the illumination people take about dimming, they're talking about duty cycle modulating the exact same current used at all duty cycles. I don't see how you do that using these LED driver chips with flyback inductors. Simplest would be a switch CCS of 30A with a shunt switch across the LED to modulate the drive.
Reply by ●April 7, 20192019-04-07
bloggs.fredbloggs.fred@gmail.com wrote...> > On Wednesday, April 3, 2019, Winfield Hill wrote: >> bloggs.fredbloggs.fred@gmail.com wrote... >>> On Wednesday, April 3, 2019, Winfield Hill wrote: >>>> >>>> What are some good switching-mode LED drivers, able >>>> to work > 26V, regulate current at 30A (ext FETs), >>>> and do fast PWM for 3000:1 or better dimming? We >>>> looked at MAX16819, what are some others you like? >>> >>> What kind of PWM circuit are you contemplating >>> for 3000:1? >> >> In addition to their current-mode control, the chips >> have a PWM input pin, and claim 2000 to 5000:1 dimming >> capability. In the fine print you discover you have >> to slow the PWM rate significantly, and also deal with >> a minimum pulse length, plus uncertainty in the on/off >> delay times. Consider, with a 20kHz PWM rate, a 100ns >> width uncertainty limits you to a 500:1 range. But if >> the PWM rate was 3kHz, you could get 3000:1 range.=20 >> To get zero, make no pulses, or to get a light level of >> 1 part in 3000, make a single 100ns pulse, every 330 us. >> >> I'm not very concerned about that spec. > > ... So when the illumination people take about dimming, > they're talking about duty cycle modulating the exact > same current used at all duty cycles. I don't see how > you do that using these LED driver chips with flyback > inductors. Simplest would be a switch CCS of 30A with > a shunt switch across the LED to modulate the drive.Correct, the MAX16819 cannot be used as a projection IC, but only for single-LED illumination. An example of what you're talking about is the MAX16818, which suggests on the front page, that you add a dimming MOSFET to short the LEDs, maintaining inductor current. I was enthusisatic about the MAX16818, but they made it NRND and substituted the MAX20078, which is as I said, NDA-required, just to see a datasheet! And it's an especially onerous NDA, you need to be a very high-volumn user, and an officer of the corporation with sufficient authority, has to sign the NDA, etc. So I'm still searching for some candidate parts, so far there's only the LT3743. This chip has worked well for us, but it's a bit of a complicated design. -- Thanks, - Win
Reply by ●April 7, 20192019-04-07
Winfield Hill wrote...> >bloggs.fredbloggs.fred@gmail.com wrote... >> >> ... So when the illumination people take about dimming, >> they're talking about duty cycle modulating the exact >> same current used at all duty cycles. I don't see how >> you do that using these LED driver chips with flyback >> inductors. Simplest would be a switch CCS of 30A with >> a shunt switch across the LED to modulate the drive. > > Correct, the MAX16819 cannot be used as a projection > IC, but only for single-LED illumination. An example > of what you're talking about is the MAX16818, which > suggests on the front page, that you add a dimming > MOSFET to short the LEDs, maintaining inductor current. > [ snip ] > So I'm still searching for some candidate parts, so > far there's only the LT3743. This chip has worked > well for us, but it's a bit of a complicated design.If the user adds an external MOSFET to short the LED, then *any* buck converter able to maintain an unchanging inductor current, despite output-voltage changes, should work well. High-side current sensing, high-frequency switching, and a fast control loop, is probably required. -- Thanks, - Win
