# LM317 -> Constant Current LED -> Transistor to boost Amperage?

#### parawizard

Joined Jul 23, 2009
10
I do projects in front projection. I am now playing with DLP projectors and LEDs with low forward voltages and high current. 3-3.5v and 9A current. I have built a small circuit using:

1 x LM317 + heatsink
1 x 1.2ohm Resistor
SST-90 LED
Power Supply 5.75v 3A

Diagram:
R1 = 1.2ohm
LED = 3Vf (up to 9A)
Vdrop over LM317 2-3.5v (haven't measured yet)
I = 1A

This lights the LED but I am looking into ways to make it much closer to a 9A circuit. Been reading and apaprently I can use some transistors to boost the current output with a LM317 in voltage regulation but I cannot find any examples using constant current. I have very limited knowledge when working with electronics. I am slowly reading the Free Circuit Textbooks. I was wondering how you would implement a Transistor into this circuit and still use constant current.

As this is a learning exercise as well as a front projection project I don't mind other suggestions but working with the parts I have here I would be hard pressed to find another Regulator such as LM338 (3A). I am looking around through my part bins though.

#### SgtWookie

Joined Jul 17, 2007
22,221
Using a linear constant current regulator for a 9A load would be terribly inefficient. You would dissipate roughly 1/2 the power in the linear regulator circuit alone. I don't believe that linear regulators are still being made for that much current.

You really need a switching-type "buck" regulator. Unfortunately, those are somewhat complex, even for experienced experimenters.

But for starters, it would help to know more about your SST-90 LEDs (attaching a datasheet would be great) and what is the color(s) of the LED(s) that you are considering.

It would also help a great deal to know if you are planning on operating multiple LEDs, and if so, can they be wired in series.

Voltage drop across an LM317:
Typically, the minimum voltage drop across an LM317 when used in voltage regulation mode is 1.7v.

There is also Vref, which is the voltage between the OUT and ADJ terminals; nominally this is 1.25v, but can be as low as 1.2v or high as 1.3v and still be within specifications.

When used in current regulation mode, the minimum voltage drop across the regulator is the original 1.7v plus Vref, so 3v is the minimum voltage drop. However, at high currents and over temperature, the 1.7v drop will vary considerably. There are graphs in the datasheet which indicate the typical voltage drop vs current and temperature.

#### SgtWookie

Joined Jul 17, 2007
22,221
I threw together an experimental switching regulator on a simulator, just so you would have something to look at. I would not suggest using this as an actual current regulator implementation.

When the circuit is first powered on, there is no voltage across Rsense, a 10m (0.01 Ohm) resistor because there is no current flow.

Because the voltage on the noninverting input of the comparator U2A is higher than that of the inverting input (that across Rsense), the output of the comparator is open (high). This causes R3 to turn on Q2, an NPN transistor, which pulls the gate of Q1, a P-channel MOSFET to ground, turning it on.

Q1 supplies current through L1, a 330uH inductor, on through Rload (your LED), and across Rsense to ground.

As current through the Q1/L1/Rload/Rsense circuit increases, the magnetic field around L1 expands, and the voltage across Rsense increases. After a certain point, the voltage on Rsense (and the inverting input of U2A) becomes higher than the noninverting input, the output of the comparator turns on, pulling the output to nearly 0v.

This turns off transistor Q2. Since Q2 is no longer sinking current, R2 pulls the gate of the MOSFET up to 10v (same as the source voltage, so Vgs=0), turning it off.

Since the current supply to L1 has been cut off, this causes the magnetic field around L1 to begin to collapse. As the magnetic field collapses, the inductor "tries" to keep the current flowing in the circuit.

D1 is a 30A Schottky diode; it provides a loop current path for L1's discharge current. C1 provides a bit of buffering for D1; in an actual circuit it may not be needed.

As the magnetic field around L1 collapses, the current through the circuit continues to flow, but decreases. After a certain point, the voltage drop across Rsense decreases to the point where the voltage on the noninverting input of U2A is once again higher than on the inverting input; the output becomes high, Q2 turns on, pulling the gate of Q1 low again, and the cycle repeats.

R7 provides a bit of hysteresis feedback. The values chosen causes the circuit to switch at a (nominal) 27kHz, just out of the range of most people's hearing.

#### Attachments

• 69 KB Views: 323

#### parawizard

Joined Jul 23, 2009
10
Using a linear constant current regulator for a 9A load would be terribly inefficient. You would dissipate roughly 1/2 the power in the linear regulator circuit alone. I don't believe that linear regulators are still being made for that much current.

You really need a switching-type "buck" regulator. Unfortunately, those are somewhat complex, even for experienced experimenters.

But for starters, it would help to know more about your SST-90 LEDs (attaching a datasheet would be great) and what is the color(s) of the LED(s) that you are considering.

It would also help a great deal to know if you are planning on operating multiple LEDs, and if so, can they be wired in series.

Voltage drop across an LM317:
Typically, the minimum voltage drop across an LM317 when used in voltage regulation mode is 1.7v.

There is also Vref, which is the voltage between the OUT and ADJ terminals; nominally this is 1.25v, but can be as low as 1.2v or high as 1.3v and still be within specifications.

When used in current regulation mode, the minimum voltage drop across the regulator is the original 1.7v plus Vref, so 3v is the minimum voltage drop. However, at high currents and over temperature, the 1.7v drop will vary considerably. There are graphs in the datasheet which indicate the typical voltage drop vs current and temperature.
Hello SgtWookie,

Thanks for the response. It would be very inefficient. I have been looking to get something done with an IC also.

National LM3409
http://cache.national.com/ds/LM/LM3409.pdf

Luminus SST-90
http://www.luminus.com/stuff/conten...3_sst_90_w_product_datasheet_illumination.pdf

Got a schematic and BoM from WEBENCH but I am having problems sourcing the parts other then the IC.

I threw together an experimental switching regulator on a simulator, just so you would have something to look at. I would not suggest using this as an actual current regulator implementation.

When the circuit is first powered on, there is no voltage across Rsense, a 10m (0.01 Ohm) resistor because there is no current flow.

Because the voltage on the noninverting input of the comparator U2A is higher than that of the inverting input (that across Rsense), the output of the comparator is open (high). This causes R3 to turn on Q2, an NPN transistor, which pulls the gate of Q1, a P-channel MOSFET to ground, turning it on.

Q1 supplies current through L1, a 330uH inductor, on through Rload (your LED), and across Rsense to ground.

As current through the Q1/L1/Rload/Rsense circuit increases, the magnetic field around L1 expands, and the voltage across Rsense increases. After a certain point, the voltage on Rsense (and the inverting input of U2A) becomes higher than the noninverting input, the output of the comparator turns on, pulling the output to nearly 0v.

This turns off transistor Q2. Since Q2 is no longer sinking current, R2 pulls the gate of the MOSFET up to 10v (same as the source voltage, so Vgs=0), turning it off.

Since the current supply to L1 has been cut off, this causes the magnetic field around L1 to begin to collapse. As the magnetic field collapses, the inductor "tries" to keep the current flowing in the circuit.

D1 is a 30A Schottky diode; it provides a loop current path for L1's discharge current. C1 provides a bit of buffering for D1; in an actual circuit it may not be needed.

As the magnetic field around L1 collapses, the current through the circuit continues to flow, but decreases. After a certain point, the voltage drop across Rsense decreases to the point where the voltage on the noninverting input of U2A is once again higher than on the inverting input; the output becomes high, Q2 turns on, pulling the gate of Q1 low again, and the cycle repeats.

R7 provides a bit of hysteresis feedback. The values chosen causes the circuit to switch at a (nominal) 27kHz, just out of the range of most people's hearing.
Thanks for the example circuit and the explanation of events.

Last edited:

#### parawizard

Joined Jul 23, 2009
10
I have decided to go with a LM3409 WEBENCH Circuit. Moved over most of the surface mount components to axial/radial so I can avoid making a PCB off the bat. I could have used copper clad I guess.

Attachments:
Schematic
Original Bill of Materials
Sourced Bill of Materials

#### Attachments

• 681.8 KB Views: 274
• 28.7 KB Views: 93
• 56.8 KB Views: 157
Last edited:

#### SgtWookie

Joined Jul 17, 2007
22,221
It looks like you found a pretty decent solution.

I'm not in the least offended that you didn't use my example; it was merely intended as an educational tool and not as a circuit to actually use. However, if you find yourself wondering why you need to use some components that were specified so particularly, you might refer back to it.

Switching power supplies are great - I'll profess that I'm an amateur at them. However, you will need to keep your circuit paths as short and as wide as possible. Long, skinny traces on a board or long component leads will have a lot of inductance, which will mess up your circuit.