Pulse-powered current limiter

Hi,

I am experimenting with white LED COBs and wanted to power them from a 
constant current source while still having the PWM dimming capability.
The current limiter should be located next to the diode and the lamp 
"interface" should be regular two wires. In other words, the current 
limiter should be capable of being powered from a +12V/open drain source 
with the PWM frequency of ~400Hz.

I started my experiments with a boring two-transistor series MOSFET/BJT 
current limiter. The voltage at the MOSFET source turns on a BJT that 
steals the MOSFET gate voltage, stabilizing the drain current. It works, 
but the V_BE tempco makes the current drift way beyond my comfort zone.
OTOH, an opamp-based current limiter + a decent reference would keep the 
current rock-solid, but it can go nuts during the PWM edge transients.

Any suggestion on how to make a limiter capable of being PWM-powered and 
maintaining stable current during the ON phase down to, say, 3% accuracy 
over the 11-15V input voltage range and 25-80 degrees C temperature 
would be appreciated. The nominal LED voltage is 9V and the required 
current is ~500mA. Are there any particularly forgiving opamps?

	Best regards, Piotr

On Fri, 18 Feb 2022 22:29:11 +0100, Piotr Wyderski
<bombald@protonmail.com> wrote:

>Hi,
>
>I am experimenting with white LED COBs and wanted to power them from a 
>constant current source while still having the PWM dimming capability.
>The current limiter should be located next to the diode and the lamp 
>"interface" should be regular two wires. In other words, the current 
>limiter should be capable of being powered from a +12V/open drain source 
>with the PWM frequency of ~400Hz.
>
>I started my experiments with a boring two-transistor series MOSFET/BJT 
>current limiter. The voltage at the MOSFET source turns on a BJT that 
>steals the MOSFET gate voltage, stabilizing the drain current. It works, 
>but the V_BE tempco makes the current drift way beyond my comfort zone.
>OTOH, an opamp-based current limiter + a decent reference would keep the 
>current rock-solid, but it can go nuts during the PWM edge transients.
>
>Any suggestion on how to make a limiter capable of being PWM-powered and 
>maintaining stable current during the ON phase down to, say, 3% accuracy 
>over the 11-15V input voltage range and 25-80 degrees C temperature 
>would be appreciated. The nominal LED voltage is 9V and the required 
>current is ~500mA. Are there any particularly forgiving opamps?
>
>	Best regards, Piotr

Depletion fet? They aren't accurate to 3%, but an LED shouldn't need
that sort of precision.

-- 

If a man will begin with certainties, he shall end with doubts, 
but if he will be content to begin with doubts he shall end in certainties.
Francis Bacon

fredag den 18. februar 2022 kl. 22.29.33 UTC+1 skrev Piotr Wyderski:
> Hi, 
> 
> I am experimenting with white LED COBs and wanted to power them from a 
> constant current source while still having the PWM dimming capability. 
> The current limiter should be located next to the diode and the lamp 
> "interface" should be regular two wires. In other words, the current 
> limiter should be capable of being powered from a +12V/open drain source 
> with the PWM frequency of ~400Hz. 
> 
> I started my experiments with a boring two-transistor series MOSFET/BJT 
> current limiter. The voltage at the MOSFET source turns on a BJT that 
> steals the MOSFET gate voltage, stabilizing the drain current. It works, 
> but the V_BE tempco makes the current drift way beyond my comfort zone. 
> OTOH, an opamp-based current limiter + a decent reference would keep the 
> current rock-solid, but it can go nuts during the PWM edge transients. 
> 
> Any suggestion on how to make a limiter capable of being PWM-powered and 
> maintaining stable current during the ON phase down to, say, 3% accuracy 
> over the 11-15V input voltage range and 25-80 degrees C temperature 
> would be appreciated. The nominal LED voltage is 9V and the required 
> current is ~500mA. Are there any particularly forgiving opamps? 

MOSFET and something like the lower voltage TL431,  TLVH431 ? 

maybe a diode and capacitor to provide a constant supply for the gate/reference

John Larkin wrote:

> Depletion fet?

I love them, but they have pretty high V_TH and it is 
temperature-dependent too. Too much voltage drop and doesn't solve the 
tempco issue -- a regular 2 BJT limiter performs better.

> They aren't accurate to 3%, but an LED shouldn't need that sort of precision.

Absolute accuracy is not required, but some value of the current should 
remain relatively stable. What's the point of PWM dimming if the lamp 
current can drift away by 20% just because the lamp has been turned on 
for some time?

Below is one idea. The differential amplifier looks good, so at the sim 
level it boils down to the tempco of the reference. Google says 5.6V 
zeners are particularly good, but I don't know how good is good. Looks 
like it would be worth torturing with a hotair gun with the Q5 and Q7 
being a BCV61 and Q1 and Q2 a BCM847.

	Best regards, Piotr

Version 4
SHEET 1 1700 1204
WIRE 0 -224 -64 -224
WIRE 160 -224 0 -224
WIRE 1360 -224 160 -224
WIRE 1536 -224 1360 -224
WIRE 0 -208 0 -224
WIRE 160 -192 160 -224
WIRE 0 -112 0 -128
WIRE 1360 -48 1360 -224
WIRE 160 -32 160 -112
WIRE 160 -32 -64 -32
WIRE 608 -32 160 -32
WIRE 160 -16 160 -32
WIRE 608 0 608 -32
WIRE 160 64 160 48
WIRE 608 96 608 80
WIRE 608 96 464 96
WIRE 608 128 608 96
WIRE 608 224 608 208
WIRE -64 400 -64 -32
WIRE 1536 416 1536 -224
WIRE 1536 544 1536 496
WIRE -64 576 -64 480
WIRE 144 576 -64 576
WIRE -64 608 -64 576
WIRE 144 608 144 576
WIRE 1360 624 1360 32
WIRE 1360 624 1248 624
WIRE 1488 624 1360 624
WIRE -240 656 -336 656
WIRE -128 656 -160 656
WIRE 288 656 208 656
WIRE 1248 656 1248 624
WIRE 1360 688 1360 624
WIRE 1536 688 1536 640
WIRE 1696 688 1536 688
WIRE 1536 704 1536 688
WIRE 144 736 144 704
WIRE 1248 736 1248 720
WIRE 1248 736 144 736
WIRE 1296 736 1248 736
WIRE 288 752 288 656
WIRE -64 800 -64 704
WIRE 32 800 -64 800
WIRE 1536 800 1536 784
WIRE 464 832 464 96
WIRE 464 832 288 832
WIRE -64 864 -64 800
WIRE 144 864 144 736
WIRE 32 912 32 800
WIRE 32 912 0 912
WIRE 80 912 32 912
WIRE 288 976 288 944
WIRE 1360 992 1360 784
WIRE -64 1008 -64 960
WIRE 144 1008 144 960
WIRE -336 1200 -336 656
WIRE 1696 1200 1696 688
WIRE 1696 1200 -336 1200
FLAG 0 -112 0
FLAG 288 976 0
FLAG 160 64 0
FLAG 1360 992 0
FLAG -64 1008 0
FLAG 144 1008 0
FLAG 1536 800 0
FLAG 608 224 0
SYMBOL pnp -128 704 M180
SYMATTR InstName Q1
SYMATTR Value BC557C
SYMBOL pnp 208 704 R180
SYMATTR InstName Q2
SYMATTR Value BC557C
SYMBOL voltage 0 -224 R0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V1
SYMATTR Value 12
SYMBOL res 272 736 R0
SYMATTR InstName R3
SYMATTR Value 1
SYMBOL voltage 288 848 R0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V2
SYMATTR Value 0.6
SYMBOL res -144 640 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R4
SYMATTR Value 1
SYMBOL npn 1296 688 R0
SYMATTR InstName Q3
SYMATTR Value BC547B
SYMBOL cap 1232 656 R0
WINDOW 3 -55 51 Left 2
SYMATTR InstName C1
SYMATTR Value 100p
SYMBOL res 1344 -64 R0
SYMATTR InstName R5
SYMATTR Value 4k7
SYMBOL zener 176 48 R180
WINDOW 0 24 64 Left 2
WINDOW 3 24 0 Left 2
SYMATTR InstName D1
SYMATTR Value BZX84C6V2L
SYMATTR Description Diode
SYMATTR Type diode
SYMBOL res 144 -208 R0
SYMATTR InstName R6
SYMATTR Value 1k
SYMBOL npn 0 864 M0
SYMATTR InstName Q5
SYMATTR Value BC547B
SYMBOL npn 80 864 R0
SYMATTR InstName Q7
SYMATTR Value BC547B
SYMBOL res -80 384 R0
SYMATTR InstName R2
SYMATTR Value 10k
SYMBOL nmos 1488 544 R0
SYMATTR InstName M1
SYMATTR Value FDS4885C_N
SYMBOL res 1520 688 R0
SYMATTR InstName R8
SYMATTR Value 1.2
SYMBOL res 1520 400 R0
SYMATTR InstName R9
SYMATTR Value 1
SYMBOL res 592 -16 R0
SYMATTR InstName R1
SYMATTR Value 9.6k
SYMBOL res 592 112 R0
SYMATTR InstName R7
SYMATTR Value 1k
TEXT -344 584 Left 2 !.tran 1

Lasse Langwadt Christensen wrote:

> MOSFET and something like the lower voltage TL431,  TLVH431 ?

I was experimenting with the MAX8515 from Mouser, but it was a total 
disaster. It looks as if the the physical part and the datasheet were 
completely unrelated. IMHO the origin excludes counterfeiting, but the 
way the part misbehaves makes me think that even the pinout cannot be 
right. If I just connect IN and PGND/GND and leave FB and OUT 
unconnected, the chip draws 40mA, while the datasheet says 1mA max. When 
I connect FB to GND, the current consumption goes to 0. The scope says 
it is not oscillating. No idea where the 40mA goes, as the high-power 
path is through the OUT pin. When connected to the FET, it jumps to the 
40mA mode and starts oscillating at the OUT. 100nF kills the 
oscillations, but the chip looks offended and keeps the FET almost off.
Same behaviour with 3 different chips.

I had similar experience last year with some Maxim boost converter -- 
the chip looked haunted, while a TI part did the job right out of the 
box. Not going to use any Maxim part anymore.

I have suddenly started liking discretes.

> maybe a diode and capacitor to provide a constant supply for the gate/reference

I considered this, but keep it a last resort solution.

	Best regards, Piotr

On Sat, 19 Feb 2022 00:28:54 +0100, Piotr Wyderski
<bombald@protonmail.com> wrote:

>John Larkin wrote:
>
>> Depletion fet?
>
>I love them, but they have pretty high V_TH and it is 
>temperature-dependent too. Too much voltage drop and doesn't solve the 
>tempco issue -- a regular 2 BJT limiter performs better.
>
>> They aren't accurate to 3%, but an LED shouldn't need that sort of precision.
>
>Absolute accuracy is not required, but some value of the current should 
>remain relatively stable. What's the point of PWM dimming if the lamp 
>current can drift away by 20% just because the lamp has been turned on 
>for some time?
>
>Below is one idea. The differential amplifier looks good, so at the sim 
>level it boils down to the tempco of the reference. Google says 5.6V 
>zeners are particularly good, but I don't know how good is good. Looks 
>like it would be worth torturing with a hotair gun with the Q5 and Q7 
>being a BCV61 and Q1 and Q2 a BCM847.
>
>	Best regards, Piotr

It's just an LED! I doubt that you would notice 20% light output
change. LEDs get dimmer as the temp goes up anyhow.

There are cheap LED current limiter chips.

Or maybe an LM1117 and a resistor.



>
>Version 4
>SHEET 1 1700 1204
>WIRE 0 -224 -64 -224
>WIRE 160 -224 0 -224
>WIRE 1360 -224 160 -224
>WIRE 1536 -224 1360 -224
>WIRE 0 -208 0 -224
>WIRE 160 -192 160 -224
>WIRE 0 -112 0 -128
>WIRE 1360 -48 1360 -224
>WIRE 160 -32 160 -112
>WIRE 160 -32 -64 -32
>WIRE 608 -32 160 -32
>WIRE 160 -16 160 -32
>WIRE 608 0 608 -32
>WIRE 160 64 160 48
>WIRE 608 96 608 80
>WIRE 608 96 464 96
>WIRE 608 128 608 96
>WIRE 608 224 608 208
>WIRE -64 400 -64 -32
>WIRE 1536 416 1536 -224
>WIRE 1536 544 1536 496
>WIRE -64 576 -64 480
>WIRE 144 576 -64 576
>WIRE -64 608 -64 576
>WIRE 144 608 144 576
>WIRE 1360 624 1360 32
>WIRE 1360 624 1248 624
>WIRE 1488 624 1360 624
>WIRE -240 656 -336 656
>WIRE -128 656 -160 656
>WIRE 288 656 208 656
>WIRE 1248 656 1248 624
>WIRE 1360 688 1360 624
>WIRE 1536 688 1536 640
>WIRE 1696 688 1536 688
>WIRE 1536 704 1536 688
>WIRE 144 736 144 704
>WIRE 1248 736 1248 720
>WIRE 1248 736 144 736
>WIRE 1296 736 1248 736
>WIRE 288 752 288 656
>WIRE -64 800 -64 704
>WIRE 32 800 -64 800
>WIRE 1536 800 1536 784
>WIRE 464 832 464 96
>WIRE 464 832 288 832
>WIRE -64 864 -64 800
>WIRE 144 864 144 736
>WIRE 32 912 32 800
>WIRE 32 912 0 912
>WIRE 80 912 32 912
>WIRE 288 976 288 944
>WIRE 1360 992 1360 784
>WIRE -64 1008 -64 960
>WIRE 144 1008 144 960
>WIRE -336 1200 -336 656
>WIRE 1696 1200 1696 688
>WIRE 1696 1200 -336 1200
>FLAG 0 -112 0
>FLAG 288 976 0
>FLAG 160 64 0
>FLAG 1360 992 0
>FLAG -64 1008 0
>FLAG 144 1008 0
>FLAG 1536 800 0
>FLAG 608 224 0
>SYMBOL pnp -128 704 M180
>SYMATTR InstName Q1
>SYMATTR Value BC557C
>SYMBOL pnp 208 704 R180
>SYMATTR InstName Q2
>SYMATTR Value BC557C
>SYMBOL voltage 0 -224 R0
>WINDOW 123 0 0 Left 0
>WINDOW 39 0 0 Left 0
>SYMATTR InstName V1
>SYMATTR Value 12
>SYMBOL res 272 736 R0
>SYMATTR InstName R3
>SYMATTR Value 1
>SYMBOL voltage 288 848 R0
>WINDOW 123 0 0 Left 0
>WINDOW 39 0 0 Left 0
>SYMATTR InstName V2
>SYMATTR Value 0.6
>SYMBOL res -144 640 R90
>WINDOW 0 0 56 VBottom 2
>WINDOW 3 32 56 VTop 2
>SYMATTR InstName R4
>SYMATTR Value 1
>SYMBOL npn 1296 688 R0
>SYMATTR InstName Q3
>SYMATTR Value BC547B
>SYMBOL cap 1232 656 R0
>WINDOW 3 -55 51 Left 2
>SYMATTR InstName C1
>SYMATTR Value 100p
>SYMBOL res 1344 -64 R0
>SYMATTR InstName R5
>SYMATTR Value 4k7
>SYMBOL zener 176 48 R180
>WINDOW 0 24 64 Left 2
>WINDOW 3 24 0 Left 2
>SYMATTR InstName D1
>SYMATTR Value BZX84C6V2L
>SYMATTR Description Diode
>SYMATTR Type diode
>SYMBOL res 144 -208 R0
>SYMATTR InstName R6
>SYMATTR Value 1k
>SYMBOL npn 0 864 M0
>SYMATTR InstName Q5
>SYMATTR Value BC547B
>SYMBOL npn 80 864 R0
>SYMATTR InstName Q7
>SYMATTR Value BC547B
>SYMBOL res -80 384 R0
>SYMATTR InstName R2
>SYMATTR Value 10k
>SYMBOL nmos 1488 544 R0
>SYMATTR InstName M1
>SYMATTR Value FDS4885C_N
>SYMBOL res 1520 688 R0
>SYMATTR InstName R8
>SYMATTR Value 1.2
>SYMBOL res 1520 400 R0
>SYMATTR InstName R9
>SYMATTR Value 1
>SYMBOL res 592 -16 R0
>SYMATTR InstName R1
>SYMATTR Value 9.6k
>SYMBOL res 592 112 R0
>SYMATTR InstName R7
>SYMATTR Value 1k
>TEXT -344 584 Left 2 !.tran 1
>
>
>
>
>
>
-- 

If a man will begin with certainties, he shall end with doubts, 
but if he will be content to begin with doubts he shall end in certainties.
Francis Bacon

On Friday, February 18, 2022 at 1:29:33 PM UTC-8, Piotr Wyderski wrote:

> I am experimenting with white LED COBs and wanted to power them from a 
> constant current source while still having the PWM dimming capability. 
> The current limiter should be located next to the diode and the lamp 
> "interface" should be regular two wires. In other words, the current 
> limiter should be capable of being powered from a +12V/open drain source 
> with the PWM frequency of ~400Hz. 

At lower voltage, there's integrated solutions for onesie white-output LEDs, like
AMC7135 (that one's about 350 mA, 6V), in three-pin packages.   At 9V nominal,
even the venerable uA723 with a  pass transistor could do
something close to valid, with its overcurrent limit; maybe 10% accuracy.

whit3rd wrote:

> At lower voltage, there's integrated solutions for onesie white-output LEDs, like
> AMC7135 (that one's about 350 mA, 6V), in three-pin packages.   At 9V nominal,
> even the venerable uA723 with a  pass transistor could do
> something close to valid, with its overcurrent limit; maybe 10% accuracy.

But would it like pulsed supply? This is the core issue, not how to 
build a limiter itself. I have no idea what's inside a real 
opamp/refrence/LDO and the models probably do no cover this aspect. Or 
maybe they do, some randomly selected LTC opamps went monkey in LTspice.

	Best regards, Potr

On 19/02/2022 7:29 am, Piotr Wyderski wrote:
> Hi,
> 
> I am experimenting with white LED COBs and wanted to power them from a 
> constant current source while still having the PWM dimming capability.
> The current limiter should be located next to the diode and the lamp 
> "interface" should be regular two wires. In other words, the current 
> limiter should be capable of being powered from a +12V/open drain source 
> with the PWM frequency of ~400Hz.
> 
> I started my experiments with a boring two-transistor series MOSFET/BJT 
> current limiter. The voltage at the MOSFET source turns on a BJT that 
> steals the MOSFET gate voltage, stabilizing the drain current. It works, 
> but the V_BE tempco makes the current drift way beyond my comfort zone.
> OTOH, an opamp-based current limiter + a decent reference would keep the 
> current rock-solid, but it can go nuts during the PWM edge transients.
> 
> Any suggestion on how to make a limiter capable of being PWM-powered and 
> maintaining stable current during the ON phase down to, say, 3% accuracy 
> over the 11-15V input voltage range and 25-80 degrees C temperature 
> would be appreciated. The nominal LED voltage is 9V and the required 
> current is ~500mA. Are there any particularly forgiving opamps?
> 
>      Best regards, Piotr
> 

A small resistor to swamp the tempco?

lørdag den 19. februar 2022 kl. 03.02.24 UTC+1 skrev David Eather:
> On 19/02/2022 7:29 am, Piotr Wyderski wrote: 
> > Hi, 
> > 
> > I am experimenting with white LED COBs and wanted to power them from a 
> > constant current source while still having the PWM dimming capability. 
> > The current limiter should be located next to the diode and the lamp 
> > "interface" should be regular two wires. In other words, the current 
> > limiter should be capable of being powered from a +12V/open drain source 
> > with the PWM frequency of ~400Hz. 
> > 
> > I started my experiments with a boring two-transistor series MOSFET/BJT 
> > current limiter. The voltage at the MOSFET source turns on a BJT that 
> > steals the MOSFET gate voltage, stabilizing the drain current. It works, 
> > but the V_BE tempco makes the current drift way beyond my comfort zone. 
> > OTOH, an opamp-based current limiter + a decent reference would keep the 
> > current rock-solid, but it can go nuts during the PWM edge transients. 
> > 
> > Any suggestion on how to make a limiter capable of being PWM-powered and 
> > maintaining stable current during the ON phase down to, say, 3% accuracy 
> > over the 11-15V input voltage range and 25-80 degrees C temperature 
> > would be appreciated. The nominal LED voltage is 9V and the required 
> > current is ~500mA. Are there any particularly forgiving opamps? 
> > 
> >     Best regards, Piotr 
> >
> A small resistor to swamp the tempco?

or an NPN with two diodes as reference, all thermally connected