Boosting LED driver

[BW]

Hi! The combination of a boosting DC/DC converter and a constant-
current regulator in a single IC seems popular for driving strings of
LEDs, especially for a low voltage input (say 12 volt) and a long
string of LEDs which need a high current. Has anybody got any
recommendations on what chip to use ? It is difficult to find these
among the dozens of similar IC's with multiple channels, dimming, too
low current or too low output-voltage etc.

So far I have one from Maxim called 16818 and a National part, LM5000
as candidates..

Best regards,

Bjorn

 

[linnix]

Hi! The combination of a boosting DC/DC converter and a constant-
current regulator in a single IC seems popular for driving strings of
LEDs, especially for a low voltage input (say 12 volt) and a long
string of LEDs which need a high current. Has anybody got any
recommendations on what chip to use ? It is difficult to find these
among the dozens of similar IC's with multiple channels, dimming, too
low current or too low output-voltage etc.

If you need the power (V & I), you should build it out of discrete
components. Power transistors need room to cool and Inductors and
capacitors are hard to build in IC. Anyway, I am using 3Q, 2C and 1L
booster circuit in an IC, but I don't need high power.

 

[BW]

If you need the power (V & I), you should build it out of discrete
components. Power transistors need room to cool and Inductors and
capacitors are hard to build in IC. Anyway, I am using 3Q, 2C and 1L
booster circuit in an IC, but I don't need high power.

Hi,

yes I need at least 500-700 mA for approx 1 ms long flashes, driving a
chain of
LEDs with a total forward voltage of maybe around 40-60 V. The
circuits I've
looked at have had the switching regulation controlled by an IC but
then used
discrete external inductors, capacitors and MOSFET's for switching the
inductor
and dimming the load. That setup sounds fine by me. I just have
trouble finding
a useable application note describing a similar setup - most
application notes
are for driving single or maybe 2-3 white LEDs at a couple of dozen mA
(obviously
since white LEDs are a major booming market!), or they are buck
regulators where
the input voltage already is 40-60 V.

I wonder if a switching integrated regulator with a switch frequency
of say 700 kHz
can ramp up and control such a short flash in a stable manner. I guess
it would, since
1 ms would be 700 switching cycles and it shouldn't take that many
cycles before it
was up enough in voltage over the output capacitor to give the needed
current.

Regards,

Bjorn

 

[linnix]

yes I need at least 500-700 mA for approx 1 ms long flashes, driving a
chain of LEDs with a total forward voltage of maybe around 40-60 V.

You would be driving several amps into the transistors.
Definitely TO-220 or even TO-3.

 

[Klaus Kragelund]

You would be driving several amps into the transistors.
Definitely TO-220 or even TO-3.

I have often needed a bucking LED driver instead. We have the typical
application in which a microcontroller controls a set of parallel LEDs
in a 7-segment 3 digit display.

The typical design has a lot of power loss in the resistors in series
with the LEDs since we have to take VF and supply tolerance into
consideration. What I would like to have is a chip that controls the
3digit 7-segment and that controls a buck converter to set the total
current in the LEDs. So if x LEDs are on, the current control is set
to x current.

Anybody know if such a device exists? It would boost the efficiency
and discard of all the resistors too.

Regards

Klaus

 

[linnix]

yes I need at least 500-700 mA for approx 1 ms long flashes,

I ran some simulations. Using 10mH inductor and 10uF capacitor, it
would take 500 ms to ram from 12V to 50V. I use base current of 8 mA
on a NPN power transistor and no output load.

Depending on your duty cycle, you would need to drive the transistor
harder.

 

[neon]

use LM317 as a current source LED have differnce voltage drop for each different diode so a current source will devide up all the voltages as required and it wil be safer for the LED is not the current that blows them up but rather heat. and all you need is one resistor or a pot if you care to vary the intensity. the device is nly good for 35v but you may offset3 regulator or that voltage

 

[Klaus Kragelund]

I ran some simulations. Using 10mH inductor and 10uF capacitor, it
would take 500 ms to ram from 12V to 50V. I use base current of 8 mA
on a NPN power transistor and no output load.

Depending on your duty cycle, you would need to drive the transistor
harder.

Why would you use such an enormous inductor?

Depending on the switching frequency, I would say something like 1uF
cap and 100uH inductor. That would allow the current to settle pretty
fast

Regards

Klaus

 

[linnix]

Why would you use such an enormous inductor?

For enormous power!

Depending on the switching frequency, I would say something like 1uF
cap and 100uH inductor. That would allow the current to settle pretty
fast

He need to store enough charges for 700mA at 50V. 100uH would take
forever to charge. My simulation model needs 500ms to pump with 10mH.

 

[James Arthur]

For enormous power!

You might think that at first, but remember--the stored energy is
(1/2)*L*i^2.

For a given peak current a smaller inductor value stores less energy,
but it can be charged proportionally faster, so you just increase the
switching frequency proportionally and there's no difference in energy
delivered.

What a lower value _does_ give you is many fewer turns on your
inductor, allowing fatter wire. Fatter wire cuts losses.

More importantly, fewer turns means you can push more current
without saturating your core, allowing you to increase the i^2 term a
bunch.

So, smaller inductors allow one to *increase* power.

Cheers,
James Arthur

 

[linnix]

You might think that at first, but remember--the stored energy is
(1/2)*L*i^2.

If the circuit is 100% efficient.

I do not think, just simulate. Using 2N2219 oscilator, 2N3055 driver,
1K R(collector), 6V VCC. 1uF forward cap, 0.1uF feedback cap and 10uF
input cap.

Assume 5K load (50V,10mA):

http://linnix.com/graph1.jpg (10mH, 14V in 300ms)
http://linnix.com/graph2.jpg (1mH, 11V in 300ms)
http://linnix.com/graph3.jpg (100uH, 7V in 300ms)

Notice that graph3 (100uH) is leveling off.

 

[James Arthur]

If the circuit is 100% efficient.

I do not think, just simulate. Using 2N2219 oscilator, 2N3055 driver,
1K R(collector), 6V VCC. 1uF forward cap, 0.1uF feedback cap and 10uF
input cap.

Assume 5K load (50V,10mA):

http://linnix.com/graph1.jpg (10mH, 14V in 300ms)http://linnix.com/graph2.jpg (1mH, 11V in 300ms)http://linnix.com/graph3.jpg(100uH, 7V in 300ms)

Notice that graph3 (100uH) is leveling off.

Adjust the frequencies so that the inductor in each circuit sees the
same peak current...

Cheers,
James Arthur

 

[linnix]

Adjust the frequencies so that the inductor in each circuit sees the
same peak current...

if the 2N2219 does not saturate or stop oscillating.

I tried 2N2904, 2N2222, 2N2219 and 2N3055, using actual spice3
models. They won't work with a small inductor.

I will work on a spice3 front end (allowing you to select the
simulation components) to prove it. I have to prove it to my customer
anyway.

 

[James Arthur]

if the 2N2219 does not saturate or stop oscillating.

I tried 2N2904, 2N2222, 2N2219 and 2N3055, using actual spice3
models. They won't work with a small inductor.

I will work on a spice3 front end (allowing you to select the
simulation components) to prove it. I have to prove it to my customer
anyway.

I should've proposed a driver too, since duty cycle matters, not just
frequency, and your 2n3055 isn't going to like being driven 100x
faster!

Now this hysteretic converter is complete crap right off the top of
my head--so don't go putting it in any airplanes--but it'll switch
fast enough to illustrate the point at hand:

=======
.-.-.-. L1 D1
| | | | 100uH MBRS360
+12 >-+-----------+------' '------+------|>|----+--------+->Vout
| | R4 | | |
| .-----------+--[100k]--+----+ | ,---'
| | | | | | --- C2 / \ ZD1
R1 [1k] | R3 [220] '----||----' | --- 1uF --- 30v
| | | C1 470pF |/ | |
.----+ |R2 +-----------------| | |
| | [1k] | |>. Q3 === ===
| |/ | |/ | ZTX1048A GND GND
'--+-----o-----| |
|>. | |>. Q2 |
Q1 | --- | 2n2222 |
2n2222| / \ === |
| --- GND |
| | D2 |
| === 1n4148 |
| GND |
'-------------------------------+
|
|
R5 [0.1]
|
|
===
GND

Cheers,
James Arthur

=========
Version 4
SHEET 1 880 680
WIRE -112 -48 -256 -48
WIRE 48 -48 -112 -48
WIRE 416 -48 48 -48
WIRE -112 0 -112 -48
WIRE 48 0 48 -48
WIRE 416 0 416 -48
WIRE -256 16 -256 -48
WIRE 176 32 144 32
WIRE 304 32 256 32
WIRE 416 112 416 80
WIRE 512 112 416 112
WIRE 624 112 576 112
WIRE 768 112 624 112
WIRE -256 128 -256 96
WIRE 144 128 144 32
WIRE 144 128 -32 128
WIRE 192 128 144 128
WIRE 304 128 304 32
WIRE 304 128 256 128
WIRE 416 128 416 112
WIRE 416 128 304 128
WIRE 416 144 416 128
WIRE -32 176 -32 128
WIRE 624 176 624 112
WIRE 768 176 768 112
WIRE 48 192 48 80
WIRE 352 192 48 192
WIRE -112 224 -112 80
WIRE -112 224 -208 224
WIRE -112 256 -112 224
WIRE 48 256 48 192
WIRE 416 288 416 240
WIRE 416 288 256 288
WIRE 624 288 624 240
WIRE 768 288 768 240
WIRE -208 304 -208 224
WIRE -176 304 -208 304
WIRE -32 304 -32 256
WIRE -32 304 -176 304
WIRE -16 304 -32 304
WIRE -32 336 -32 304
WIRE 416 352 416 288
WIRE 48 400 48 352
WIRE -32 416 -32 400
WIRE -112 448 -112 352
WIRE 256 448 256 288
WIRE 256 448 -112 448
WIRE 416 480 416 432
FLAG 48 400 0
FLAG 416 480 0
FLAG -256 128 0
FLAG 624 288 0
FLAG -32 416 0
FLAG 768 288 0
SYMBOL npn -16 256 R0
SYMATTR InstName Q1
SYMATTR Value 2N2222
SYMBOL npn -176 256 R0
SYMATTR InstName Q2
SYMATTR Value 2N2222
SYMBOL npn 352 144 R0
SYMATTR InstName Q3
SYMATTR Value ZTX1048A
SYMBOL res 400 336 R0
SYMATTR InstName R5
SYMATTR Value .1
SYMBOL res 32 -16 R0
SYMATTR InstName R3
SYMATTR Value 220
SYMBOL ind 400 -16 R0
SYMATTR InstName L1
SYMATTR Value 100µH
SYMBOL res -128 -16 R0
SYMATTR InstName R1
SYMATTR Value 1k
SYMBOL voltage -256 0 R0
SYMATTR InstName V1
SYMATTR Value 12v
SYMBOL schottky 512 128 R270
WINDOW 0 32 32 VTop 0
WINDOW 3 0 32 VBottom 0
SYMATTR InstName D1
SYMATTR Value MBRS360
SYMATTR Description Diode
SYMATTR Type diode
SYMBOL cap 608 176 R0
SYMATTR InstName C1
SYMATTR Value 10µF
SYMBOL cap 256 112 R90
WINDOW 0 0 32 VBottom 0
WINDOW 3 32 32 VTop 0
SYMATTR InstName C2
SYMATTR Value 470pF
SYMBOL diode -16 400 R180
WINDOW 0 24 72 Left 0
WINDOW 3 24 0 Left 0
SYMATTR InstName D2
SYMATTR Value 1N4148
SYMBOL res 272 16 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R4
SYMATTR Value 100k
SYMBOL res -48 160 R0
SYMATTR InstName R2
SYMATTR Value 1k
SYMBOL zener 784 240 R180
WINDOW 0 24 72 Left 0
WINDOW 3 24 0 Left 0
SYMATTR InstName D3
SYMATTR Value DFLZ33
SYMATTR Description Diode
SYMATTR Type diode
TEXT -290 506 Left 0 !.tran 0 100mS 0 1uS

 

[linnix]

Yes, the ZTX1048A works better, but only if i use 330uH. It boosts
from 6V to 30V in 300ms. R9 is the output load of 15mA.

http://linnix.com/graph4.jpg

vcc 1 0 6
q1 4 2 0 2n2219
q2 7 5 0 ZTX1048A
r1 1 2 2700
r2 1 4 330
r3 1 5 47
r9 8 0 2k
c1 1 0 1u ic=0v
c2 7 2 1u ic=0v
c3 4 5 500n ic=0v
c4 8 0 100u ic=0v
l1 1 7 330u

 

[linnix]

Yes, the ZTX1048A works better, but only if i use 330uH. It boosts
from 6V to 30V in 300ms. R9 is the output load of 15mA.

http://linnix.com/graph4.jpg

vcc 1 0 6
q1 4 2 0 2n2219
q2 7 5 0 ZTX1048A
r1 1 2 2700
r2 1 4 330
r3 1 5 47
r9 8 0 2k
c1 1 0 1u ic=0v
c2 7 2 1u ic=0v
c3 4 5 500n ic=0v
c4 8 0 100u ic=0v
l1 1 7 330u

PS: I don't know if this would be a problem or not. The ZTX1048A
Collector to Emitter break down voltage is only 17.5V at 10mA. How
would you model this in spice?

 

[Don Klipstein]

PS: I don't know if this would be a problem or not. The ZTX1048A
Collector to Emitter break down voltage is only 17.5V at 10mA. How
would you model this in spice?

I have only a bit of basic knowledge of Spice due to doing more work
that is better done than simulated.

I suspect a reduced C-E breakdown voltage as a function of current (as
in higher) is due to "forward bias second breakdown". This is a
phenomenon of bipolar transistors where power dissipation capability is
reduced as a result of higher C-E voltage, even if such higher C-E voltage
is well within the device's C-E forward bias voltage limit.

In my past experience of reading power transistor datasheets, power
dissipation capability tends to be impaired by "forward bias second
breakdown" when C-E voltage is above something like 30-50 volts (varies
somewhat from one device to another). This gets "less bad" when the
bipolar transistor in question has "hometaxial" structure rather than
"epitaxial", though "hometaxial" has appeared to me less-desirable by
being slower and "epitaxial" is more common.

The forward-bias second-breakdown phenomenon also has tolerance for
higher instantaneous power dissipation at "offending" C-E voltages if this
has to be withstood for pulses (generally a few milliseconds or less,
better still if fractional millisecond or several microseconds) at a low
duty cycle. Transistors with "forward bias second breakdown"
vulnerability do often do well as switchers since in such duty they tend
to only need to dissipate high power with high C-E voltage for only
microsecond ballpark with low duty cycle, and otherwise do their best at
approximating an open or a short.

- Don Klipstein ([email protected])

 

[linnix]

I have only a bit of basic knowledge of Spice due to doing more work
that is better done than simulated.

I suspect a reduced C-E breakdown voltage as a function of current (as
in higher) is due to "forward bias second breakdown". This is a
phenomenon of bipolar transistors where power dissipation capability is
reduced as a result of higher C-E voltage, even if such higher C-E voltage
is well within the device's C-E forward bias voltage limit.

In my past experience of reading power transistor datasheets, power
dissipation capability tends to be impaired by "forward bias second
breakdown" when C-E voltage is above something like 30-50 volts (varies
somewhat from one device to another).

Breakdown of 30 to 50 V is more typical. But the ZTX1048A will
breakdown as low as 17.5V, which is not a problem for my need (6V to
15V). However, it will indeed be a problem beyond that. Anyway, I
would rather use a 2N2222A with higher breakdown voltage.
Unfortunately, neither one is available in SOT-23, so back to the
drawing (layout) board. My current board is using BC847B, which peak
at 12V in simulations.

 

[James Arthur]

PS: I don't know if this would be a problem or not. The ZTX1048A
Collector to Emitter break down voltage is only 17.5V at 10mA. How
would you model this in spice?

Yes, that's absolutely a goof in real life, but of no consequence in
LTSpice--it doesn't care. I've also supplied Q3 with 50mA of base
drive for a <1A collector current, which is absurdly much for a
ZTX1048, but possibly necessary for a clunkier (lower gain)
transistor.

Your listing seems to have left out the rectifier diode, I assume from
Q2(c) to C4 and R4 at node 8? And your power supply is 6v, right? (I
used +12v for mine because the OP asked for it.)

I see a few problems with your netlist:
-the circuit is not guaranteed to start. Q1 can get stuck saturated
'ON', Q2 winds up stuck 'ON' as well, and oscillation fails.
-I'm not sure how much current you want from this, but the timing is
set for very heavy currents, at least in the file above. I assume
there's been some mistake.

-From +12v my circuit produces 33v across 10uF in 1.7mS (120mA @ 33v
= 4W), so we may be thinking different animals here... If you only
need 15mA, then a 2n2222a may be fine for your Q2.

Anyway, my gizmo tolerates inductor changes easily--which is
convenient for exploring the question of power delivered vs: inductor
value--should you care to play.

Best wishes,
James Arthur

 

[James Arthur]

Anyway, I
would rather use a 2N2222A with higher breakdown voltage.
Unfortunately, neither one is available in SOT-23, so back to the
drawing (layout) board.

The 2n2222a in SOT-23 is called a MMBT2222A and is marked '1P' on its
case. They're everywhere.

Zetex has a number of higher-voltage high-current transistors e.g.:
http://www.zetex.com/3.0/3-3-2b.asp?rid=5

but if all you want is 15mA you might as well use the MMBT2222a.

Cheers,
James Arthur