Problems with constant current switching supply at low current settings

[Tim Mitchell]

Hello, I've designed a constant current switching supply using a UC3843.
The current needs to be variable from 350mA down to zero to drive a
stack of Luxeon leds from a 48V supply. I'm doing this by adding a
voltage offset to the feedback loop. The feedback is from a 3R3 current
sense resistor & differential amp.

The problem: I can't get the UC3843 to produce a current less than about
10mA. It just turns off at this point. The LEDs are still quite bright
at 10mA and the "snap" off is noticeable. I am not sure why the device
will not run at a lower duty cycle to give a lower current. I've tried
several different configurations of the circuit but the device behaves
the same, making me wonder if there is a minimum duty cycle below which
it won't operate. Does anyone have any advice?

 

[Harry Dellamano]

Hello, I've designed a constant current switching supply using a UC3843.
The current needs to be variable from 350mA down to zero to drive a
stack of Luxeon leds from a 48V supply. I'm doing this by adding a
voltage offset to the feedback loop. The feedback is from a 3R3 current
sense resistor & differential amp.

The problem: I can't get the UC3843 to produce a current less than about
10mA. It just turns off at this point. The LEDs are still quite bright
at 10mA and the "snap" off is noticeable. I am not sure why the device
will not run at a lower duty cycle to give a lower current. I've tried
several different configurations of the circuit but the device behaves
the same, making me wonder if there is a minimum duty cycle below which
it won't operate. Does anyone have any advice?

Dimming would be easy in voltage mode control instead of current mode.
Throw out that diff-amp and sense the 48V for feedback. Dimming is just a
pot from the 5V ref to the error amp input. This will also give you constant
intensity's over temp. Sounds like a stop light array.
Let the keyboard spanking begin!
Harry

 

[Tim Mitchell]

Dimming would be easy in voltage mode control instead of current mode.
Throw out that diff-amp and sense the 48V for feedback. Dimming is just a
pot from the 5V ref to the error amp input. This will also give you constant
intensity's over temp. Sounds like a stop light array.

No, it's a lighting product. The reason for current mode control is that
the LED drive current has to be closely regulated to 350mA max.

 

[Harry Dellamano]

No, it's a lighting product. The reason for current mode control is that
the LED drive current has to be closely regulated to 350mA max.

Tim, drive your array with an external DC source (current limited source is
easiest) and measure the exact current, voltage and light intensity to allow
dimming from 0 to 100%. Take sets of data at ambient temperatures that this
unit will operate in. Do not include any current limiting resistors inside
the array being measured. Let's see that data.

Harry

 

[Tim Mitchell]

Tim, drive your array with an external DC source (current limited source is
easiest) and measure the exact current, voltage and light intensity to allow
dimming from 0 to 100%. Take sets of data at ambient temperatures that this
unit will operate in. Do not include any current limiting resistors inside
the array being measured. Let's see that data.

mmmm... I see where you're going, but there are additional problems I
haven't mentioned yet. The array is not a fixed number of LEDs. The same
supply unit must be able to drive a series connected stack of between 4
and 12 LEDs. Also, different coloured LEDs with different forward
voltages are involved. This is why I've gone for a switching approach
rather than linear, and why I am looking to regulate the current rather
than the voltage.

 

[Harry Dellamano]

mmmm... I see where you're going, but there are additional problems I
haven't mentioned yet. The array is not a fixed number of LEDs. The same
supply unit must be able to drive a series connected stack of between 4
and 12 LEDs. Also, different coloured LEDs with different forward
voltages are involved. This is why I've gone for a switching approach
rather than linear, and why I am looking to regulate the current rather
than the voltage.

Use you exact same switching regulator, change the voltage feedback ratio
for different arrays. Adding power factor correction is also easy with this
single stage. Let's see that data before J.F. gets all over me again.
Piece of cake!

 

[Genome]

Hello, I've designed a constant current switching supply using a UC3843.
The current needs to be variable from 350mA down to zero to drive a
stack of Luxeon leds from a 48V supply. I'm doing this by adding a
voltage offset to the feedback loop. The feedback is from a 3R3 current
sense resistor & differential amp.

The problem: I can't get the UC3843 to produce a current less than about
10mA. It just turns off at this point. The LEDs are still quite bright
at 10mA and the "snap" off is noticeable. I am not sure why the device
will not run at a lower duty cycle to give a lower current. I've tried
several different configurations of the circuit but the device behaves
the same, making me wonder if there is a minimum duty cycle below which
it won't operate. Does anyone have any advice?

Hmmm... Unless you have some nifty sort of inverting drive thing going up to
a P-channel mosfet then you can't be using a buck configuration. You also
can't be using a standard boost because the diodes would be biased on by the
DC supply.

That leaves you with some form of isolated converter or it's a flyback. That
is the inductor is in the input supply line, a n N-channel mosfet pulls its
other end down and then releases it under control of the UC3843, it flys
back and charges up a filter capacitor via a rectifiying diode which drives
the LED(s) It sort of looks like the LED(s)/Filter cap and rectifier diodes
are in parallel with the inductor.

If that's the case then your current sense resistor sits (should sit) in
series with the LED(s)/Filter cap and the input supply and then you use your
diff amp to level shift the signal down to the UC3843 error amplifier.
Whilst still using a current sense resistor in the mosfet source for
feedback to the ILIM pin.

There may be something about the internals of the UC3843 that means it can't
produce a zero pulse width if you don't inject an additional bit of an
offset into the ILIM pin.

Whilst the data sheet, once upon a time, claimed somewhere that the PWM
latch was reset dominant.... if there is no signal to reset it, the mosfet
is off, then you will get a minimum pulse width, to turn the mosfet on, of
the comparator delay.

Therin may lie your problem.

DNA

 

[legg]

There may be something about the internals of the UC3843 that means it can't
produce a zero pulse width if you don't inject an additional bit of an
offset into the ILIM pin.

Whilst the data sheet, once upon a time, claimed somewhere that the PWM
latch was reset dominant.... if there is no signal to reset it, the mosfet
is off, then you will get a minimum pulse width, to turn the mosfet on, of
the comparator delay.

The 3843 should cycle-skip quite nicely, if it's not supplying it's
own juice ( in which case hiccouphing would be fairly obvious).

There's no guarantee that the luminosity of the led is in any way
linear at the lower end of the operating current, partucularly if the
peak to average value changes simultaneously (as cycle skipping can do
into a constant voltage load).

At the top end, 350mA continuous sounds like a lot of overdrive for an
LED, if'n yer askin' me.

RL

 

[legg]

If that's the case then your current sense resistor sits (should sit) in
series with the LED(s)/Filter cap and the input supply and then you use your
diff amp to level shift the signal down to the UC3843 error amplifier.
Whilst still using a current sense resistor in the mosfet source for
feedback to the ILIM pin.

There may be something about the internals of the UC3843 that means it can't
produce a zero pulse width if you don't inject an additional bit of an
offset into the ILIM pin.

There also the likelihood that converter noise is introducing a
hysteritic effect into the analog control stage.

RL

 

[Tim Mitchell]

Use you exact same switching regulator, change the voltage feedback ratio
for different arrays. Adding power factor correction is also easy with this
single stage. Let's see that data before J.F. gets all over me again.

No, the user buys this box into which they can plug between 4-12 LEDs.
The unit has to be able to work without the user having to mess with it.

The data you are asking for will not help, you have to drive luxeon LEDs
in a constant current circuit. These are not normal LEDs.

 

[Tim Mitchell]

The 3843 should cycle-skip quite nicely, if it's not supplying it's
own juice ( in which case hiccouphing would be fairly obvious).

There's no guarantee that the luminosity of the led is in any way
linear at the lower end of the operating current, partucularly if the
peak to average value changes simultaneously (as cycle skipping can do
into a constant voltage load).

At the top end, 350mA continuous sounds like a lot of overdrive for an
LED, if'n yer askin' me.

You've not played with the Luxeon LED's then? The 5 watt versions
require a drive of 750mA. The 1 watt versions I'm using have a
recommended drive of 350mA. These are not normal LED's!
see http://www.lumileds.com for details.

No, the luminosity of the LED is not linear with current, but I can
handle that in the drive software.

I think there is some timing issue with the PWM output of the UC3843 and
I will try experimenting with the current limit pin.

 

[R.Legg]

You've not played with the Luxeon LED's then? The 5 watt versions
require a drive of 750mA. The 1 watt versions I'm using have a
recommended drive of 350mA. These are not normal LED's!
see http://www.lumileds.com for details.

Not likely to. I saw some local work on this some years ago for
automotive, traffic and advertising, but I see applications are now
already assembled off-shore.

No, the luminosity of the LED is not linear with current, but I can
handle that in the drive software.

Unless you're dealing with a negative resistance region - not likely
for an entire string, but perhaps just one bad one shuts the others
off. Depends on the amplitude of the snap-off/on voltage.

http://www.lumileds.com/pdfs/protected/AB22.PDF

I think there is some timing issue with the PWM output of the UC3843 and
I will try experimenting with the current limit pin.

Check for stability, slow it down. Don't use your eyes as the only
measuring instrument. See if another manually-controlled linear source
gets the same effect.

RL