Help required designing a DC-DC buck constant current power supply

[m4rv]

Hi All,

I'm currently working on a project and I need to design 3 separate constant current switch-mode power supplies to drive some laser diodes. I will need to create 4 separate supplies to power 1 Blue, 1 Green and 2 Red diodes.

The voltage drop and current handling for the diodes are:
Blue: 4.8v - 1.5A
Green: 4.7v - 1.6A
Red: 2.3v - 870mA

The diodes were supplied with a driver for each one but they are really inefficient (<50%) and so get pretty hot.
The whole project will be mounted on a single large heatsink with appropriate cooling and the supplied drivers aren't physically laid out to be mounted this way. I will be designing a single PCB for the 4 supplies together, taking into account the heatsink mounting.

Another requirement is that the brightness must be controllable from a micro controller PWM output, either by directly modulating the current, or filtered to a constant voltage and used as part of the feedback circuit.

My power supply design requirements:

• Regulated 12v DC input
• Constant current output, up to 1.6A
• PWM or voltage controlled current output
• Controller IC readily available in low quantities in the UK, from RS/Farnell etc
• Controller IC must be hand solderable without specialist tools/equipment (i.e. not a no-lead package, or hidden "power pads")[\li]
  

I understand the majority of the theory behind how SMPS work (Daves "lab power supply design" videos certainly helped a lot) but I have no practical experience building them. I'm finding it difficult to get any information as to how to select the main components for my specific application. There seem to be so many different options of controller ICs available with different specifications but I cant work out which specs I should really be taking into account.
I have found a few examples of circuits which achieve almost what I need but they all fall short somehow. E.g. Not enough output current, no current control, utilising parts which don't exist any more or parts I can't find unless I'm buying 2500 of them.

Some of the questions I have in my head, specifically for my application:

• What sort of switching frequency should I be looking at
• Can I use pretty much any controller IC in a constant current configuration, or should I be looking for a specific current mode device
• How do I select my switching device? Should I be looking at a mosfet or BJT based design?
• ...I'm sure there are more but my brain is a bit fried from looking at datasheets and searching for info .

One device I've just come across searching RS website is the LM25085 from TI.
Datasheet: www.ti.com/lit/ds/symlink/lm25085.pdf.
Would this device be suitable for a constant current configuration? Could I just modify the "typical application" circuit to use a current sense resistor, error amplifier, etc? Does the "Constant on-time" principle make any difference to my application?

Anyway, I would greatly appreciate any pointers or advice that anyone can give me. I'm lost in a sea of information which I don't fully understand or know what to do with.

Thanks for your time!

[Richard Head]

I wonder how important cost is in this application?
If cost isn't too critical and if galvanic isolation isn't required I would think of two multi-phase synchronous buck converters (LM5119 or LT3845A) configured for dual outputs each. That'll give you four independent controllable DC outputs with just two tiny IC's. Instead of the voltage feedback from the classic potential divider network I'd implement a constant current network using an op-amp with a current sense resistor. You can feed your LED setpoint current to each op-amp from your micro's DAC. It's probably the smallest solution in terms of board real estate, but not the cheapest due to the cost of each controller.
Alternatively you could opt for four UC3844's in a buck configuration without the synchronous rectifiers (just a 1N5820 or similar Shottkey) at the expense of efficiency.
A third option is to use a single buck converter to drop the 12V supply to about 6V (regulated) and then do simple linear regulators for each LED. Not quite as efficient as the first proposal but probably adequate unless efficiency is critical.
So many ways to skin a cat!

[m4rv]

Hi Richard,

Thanks for your reply.

There are many ways to accomplish this, I'm sure!

Cost is somewhat important, we are only producing a small number of the products (for our own use only, not for resale) but we would still like to keep costs down where we can.

Looking at the first two controllers you suggested, I would say that the costs are a little to much to be spending on one device.

I think the UC3844 might be the way to go. I haven't seen it when I've been searching on suppliers websites, I think because it comes under the category of "PWM controllers" and I've only been looking in "DC-DC Controllers". I don't think I understand the difference. For example, the MC34063 falls under the DC-DC Controller category and I cant really see THAT much difference between it and the 3844 (apart from the current mode capability). If someone could explain the difference, I'd appreciate it.

So, with regards to the PWM control (my micro doesn't have an actual DAC) can I modulate either the input, or output of my supply with another mosfet for example? Or is it best to adjust the current on the sense pin of the controller?

Also, I'm just looking through the datasheet for the UC3844, it says the output "is capable of up to ±1.0 A peak drive current". Why does it need to be able to handle such a large current when its designed to be connected to a MOSFET. I thought that very little current flowed from/to a gate pin.

[Richard Head]

m4rv
I have re-read my post and it gives the impression that no external MOSFET's are required!
Indeed they are, but they can be small, such as SO-8 MOSFET's or maybe even SOT-23's.
The UC3844 (available in SO-8) is a very common current mode controller that is dirt cheap and readily available. If used with an external MOSFET and SMD inductor can yield a simple buck converter. The output current figure quoted (+/-1A) is the peak current that the device can supply to the gate of the MOSFET. The MOSFET requires a substantial current to rapidly charge and discharge its input capacitance when switched on or off. Once the MOSFET is on or off it requires no current at all, as you say. This current (i=Cdv/dt) needs to gets higher as the switching frequencies increase. When switching frequencies approach 500khz and above the gate drive circuitry becomes quite beefy.
The MC34063 is a PWM controller and the UC3844 is a current mode controller. There is a difference. Whenever possible try to use current mode control unless there's a good reason not to. Current mode control make control loop compensation a little easier and protects the power switch under transient conditions.
So, with regards to the PWM control (my micro doesn't have an actual DAC) can I modulate either the input, or output of my supply with another mosfet for example? Or is it best to adjust the current on the sense pin of the controller?
It will be best to control the LED current by means of wiring the controller as a constant current source for the LED. This will require an external op-amp and current sense resistor. The output of the op-amp must go to the feedback pin of the controller. That means that each LED in enclosed in a control loop that regulates the current to the value that you choose. If the supply voltage or LED temp changes the current will remain constant.

[m4rv]

Hi Richard,

Ah, no, your post was fairly clear to me! I always planned that I would need an external switch of some kind for output currents like this. Thanks for the clarification though.

The gate capacitance makes sense.

Another point that I should have put in the specification of the power supply is that it must be able to vary the current delivered to the laser diodes all the way down to 0. The application for the project will require the diodes to be off for long periods of time with the unit sitting idle until they are turned on in an instant. Do i need to make any special considerations for this or will I be able to achieve it just by using the current sense pin?

When it comes to selecting the correct switching device, I'm assuming that the drain-source current handling needs to be somewhat larger than the maximum current I need to output? How do I go about calculating this?

Thanks.

[Richard Head]

m4rv
Do i need to make any special considerations for this or will I be able to achieve it just by using the current sense pin?
Just set the output current to zero as you suggest. No problem.
For the buck topology (continuous inductor current mode) the peak MOSFET switch current is the inductor current plus half the inductor ripple current (small assumption). The average MOSFET current is about half that, but depends on the duty cycle of the switch. The average current is used to calculate the MOSFET dissipation.
When selecting the MOSFET you generally want to balance the on state losses with the switching losses. To calculate the on state losses: P= i^2Rdson x (duty cycle)  ; Rdson = Rdson at working temp!
To calculate switching losses: P = 0.5 VIf(ton+toff)   ; ton and toff = MOSFET transition times
Total MOSFET losses is the sum of the two. (small assumption)
When it comes to selecting the correct switching device, I'm assuming that the drain-source current handling needs to be somewhat larger than the maximum current I need to output? How do I go about calculating this?
The max current that the switching device can handle is normally never reached and the device will be dissipation limited long before it reaches its max current. That's what's nice about current mode control: You can be certain that the peak switch current will never be exceeded, even under unusual transient conditions.
The switch dissipation calc is an integral part of the design and must be performed. For that you need to know the junction-case thermal resistance and case-sink thermal resistance as well as max ambient temp. First step is to select a device that looks suitable and do the calcs. If it's a bad choice the calcs will show it as some parameter being exceeded.

[elecom]

Years passed. Is there any progress about your project?