Inductor for buck SMPS.

[treez]

Hello,

I have a 24V DC bus from which i need to drive a load which is 18V, 200mA. (the load is a series parallel bank of LEDs)

Therefore, i am going to use a Buck SMPS. The ripple requirement and transient response requirements are not that strict. -The key to this design is that it needs to use as few components as possible.

Also, i don't want the switching frequency to be much more than 25KHz.

Anyway, since i don't want to use MOSFET drivers (its extra components) i want to just use a PNP power darlington as the buck's switch transistor.

I need an off-the-shelf inductor of value ~1.2mH (or higher)and for it not to saturate until above 450mA. There seem to be none around with this rating. I would be very grateful if any reader has ideas where i could get such a part.

[The average inductor current is 200mA, -i want the peak to be ~350mA, MAX_on_time will be about 30us, therefore my requirement is for about 1.2mH)]

I cannot use an off-the-shelf SMPS for this, since there are several similar loads on the 24V DC bus, and they will all need to be switched "out-of-phase" with one another to reduce overall pulse currents on the 24V DC bus. -Due to this switching strategy, i must use a microcontroller as the oscillator for all the SMPS's and hance can't use an off-the-shelf SMPS as they would all have their own oscillators.

Amy ideas for my inductor "hunt" would be greatly appreciated. Once again, it's ~1.2mH (or higher), and saturation current above ~450mA.

 

[MacGyverS2000]

Your requirements sound arbitrary... what is all of this for? Will there be 10 made or 1 million? Your answer will determine what advice we give. Your component requirements don't appear to have much sound engineering behind them... some values don't even match up (such as your max on time versus your desired SMPS frequency).

For such a low current level, small change in input-to-output voltage, and your requirement for low component count, why not simply use a linear regulator?

Dan - Owner

http://www.Hi-TecDesigns.com

 

[OperaHouse]

For one of applications an old computer power supply might be a good source for the inductor. An analog regulator like a LM317 can be used as a switcher. These will find their own resonance no matter what inductor you use. Even an old filament transformer can be used. Lots of inductors on ebay and you can wind your own toroid. I just don't see a problem here.

 

[itsmoked]

Try here.

https://www.amidoncorp.com/

If you have so many LEDs why don't you just sting them all in series and use a single transistor to turn them ON and OFF?

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[VEBill]

Further to what Keith suggested, can't you just 'duty cycle' the LEDs and use the higher voltage?

 

[itsmoked]

Absolutely!

They only need to maintain an AVERAGE current. They have a fairly high peak current allowance.(Like 10X !)

No one will be able to even see a 25kHz blink.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[DHambley]

There's a lot of inductors at DigiKey. For example a JW-Miller 5900-222-RC 1.5mH at 0.5Amp is $4.00.

 

[MacGyverS2000]

This sounds like a hobby project, not a professional design...

Dan - Owner

http://www.Hi-TecDesigns.com

 

[GonzaloEE]

Smells like homework
Reminds me my old days when I winded my own toroid for a 15KHz (whistling) SMPS.

Good Luck, anyway!
Gonzalo

 

[jimkirk]

I'm guessing you really want a current source to put out 200 mA and the LED string will have whatever voltage it needs to do that, which apparently will be around 18 volts.

For "as simple as possible" for this application, why even use a switcher? A 200 mA current limiter operating off the 24 volts can be done with one transistor (probably something in a TO-220 for through hole or Dpak or D2pak for surface mount), an LED as a reference voltage (green works well), an emitter resistor to set the current I=(Vdiode-Vbe)/R and another resistor to bias the LED. Four components. The efficiency will be about 75% which isn't too bad, probably not much worse than the switcher you're contemplating (control circuitry, Darligton saturation voltage & turn off speed, losses in te 1.2 mH inductor...). With the linear circuit the transistor will dissipate just over one watt, and absolutely no switching noise.

 

[treez]

Hello,

..Thankyou for these excellent replies....

Background...

We are an LED company in Maryland. We sell thousands of LED lamps per year for architectural lighting.

Currently we use linear regulators to adjust the LED brightness.

However, we want to move (-but move slowly, "bit_by_bit") to a more efficient solution where we use pulsing of the LEDs to adjust brightness....and drop MINIMUM volts across any series resistors..preferably not having a series resistor at all)

We are worried that LEDs in pulsed mode will not last as long as LEDs with a constant variable current through them.

-It is for this reason that we want to move slowly...first of all, we are just taking one 24V, (12W) LED lamp, and converting that to pulsed operation….and we will monitor the MTBF (time till failure).

The point is, we don’t (yet) know if it will all work out…so we want to keep most of our current electrical and mechanical fixtures and design an initial pulsed_LED product that is compatible with our presently existing fixtures…..so that we can quickly go back to our linearly regulated design if pulsing gives too many LED failures.

…since the LED PCB for this product drops 18 volts across the LEDs at full brightness…..we need the switcher(s) to give it that voltage. (we don’t yet want to change this LED PCB).

We assemble large quantities, so an off_the_shelf inductor is needed to cut assembly time. (don’t want our people winding inductors).

Currently, these products need a fan when LEDs are on full brightness…and that takes a few Watts that we’d like to cut down....hopefully this more efficient pulsed_LED idea will allow us to downsize the fan....even operate it in pulsed mode using a microcontroller so it's average current draw is even less.

All these LED lamps run off our power supply box…the point is, the more efficient we make the LED lamps, the more we can sell per Power supply box. ($$$$$)

We already have a low component count SMPS controller for the buck converters, -for which the oscillator is a microcontroller. (it gives the phase shifted oscillations to each buck to reduce pulse currents on the 24V bus.).

Some of our products are completely enclosed, have no fan, and are used outdoors in places where it’s hot even at night……if we can get a good pulsed LED dimmer for use here then we’ll be doing well, as it will run cooler.

In fact, with the fan ON, the linearly regulated LED lamp under investigation is only about 50% efficient on full brightness.

I don’t want an all series LED solution due to its total failure mode.

In reply to above…..The (24 TO 18V) Buck is switching at 25KHz, so T = 40us and max on time will be about ~30us. (Vout = DVin). I was surprised to hear that these figures don't match up.

I was hoping I would be able to get away with a PNP Power Darlington for the Buck switch transistor, as it can be driven on without having to use a pricey high current driver.....
In fact, i may just end up using a PNP (non Darlington) power BJT and driving it from the microcontroller by putting the micro output into a small NPN in Common Enmitter mode, and then connecting the PNP base to a divider in the collector of the NPN, which will turn on the Power PNP when the NPN turns ON.......
...I can speed up switch-OFF time of the power PNP by putting a fast diode across the Collector-Base junction of the PNP and the same for the NPN (Cathode to collector). This speeds up switch-OFF by sweeping out the carriers in the base-collector junction. -I'm not immediately quite sure how but my scope seems to tell me it does!

By the way, the Dual, anti-phase switched Bucks use just 4 IC's between them....

1 8 pin PIC microcontroller
1 dual NAND latch's
1 Dual comparator IC
1 AND gate IC

...The SMPS controller configuration is like a UCC28C42 current mode controller but with quite a lot cut out.

 

[VEBill]

Once upon a time I once drove a rental car (GM) where the dashboard had many LEDs. Every time that I swiveled my head while driving at night, the dashboard LEDs left a trail of dots in my peripheral vision. It was extremely distracting. Most cars don't have this issue (switching frequency too low?), just the badly designed ones.

Point is that you need to beware of switching LEDs. Too low and they are visually distracting. Too high and you'll get QSL (reception report) post cards from distant shortwave listeners.

You can use a switch (very efficient) to duty cycle the power to set an average current (filtered with a suitably large capacitor and perhaps some series filtering if required).

The expensive, precious, LEDs will see nothing but pure DC current.

The higher the frequency, the less capacitance required. The switching noise (EMI) never leaves the PS box (another important issue).

Switching is used ALL THE TIME to make pure DC. Probably 99% of the PS market. You can control and regulate current almost as easily as voltage. You just feedback a current sample instead of a voltage sample.

I still don't see any requirement for a coil. I'd use nothing but a big FET (as well as all the control circuit).

 

[MacGyverS2000]

treez, where are you located? I'm in MD myself, about halfway between Baltimore and D.C.


I'll think on this a bit and post more useful info when I'm not tied up at work...

Dan - Owner

http://www.Hi-TecDesigns.com

 

[GS3]

I cannot use an off-the-shelf SMPS for this, since there are several similar loads on the 24V DC bus, and they will all need to be switched "out-of-phase" with one another to reduce overall pulse currents on the 24V DC bus. -Due to this switching strategy, i must use a microcontroller as the oscillator for all the SMPS's and hance can't use an off-the-shelf SMPS as they would all have their own oscillators.

I do not understand this. All it takes is a capacitor at the switch input. Using a microcontroller seems much more complex than necessary. If you want low input and output ripple and minimal RFI then Cuk topology seems a good choice.

 

[jimkirk]

Thanks for the additional information on your application. Have you looked at Microchip's MCP1630? It's a Pulse Width Modulator designed to built switch mode power supplies with a clock supplied by a microprocessor, so you have complete control of that, as many phases as you want, and the clock can be used to limit the duty cycle of the supply so you can do soft starts and such. They operate up to 1 MHz, and drive MOSFETs directly, no extra parts needed for that. Easily dim-able, and properly designed there won't be much in the way of pulses of current, just DC with some controlled amount of ripple. Their website has many application notes, including LED drivers, and many evaluation boards are available. I've built LED drivers with them, driving 6 series white LEDs from a lithium battery.

Why, exactly, do you want to limit the switching frequency to 25 kHz? The 1.2 mH inductor is going to be larger and much more expensive than the 10 uH or so inductor you could use at several hundred kHz. And a good MOSFET will get you better efficiency than a Darlington.

Most designs like this will use N channel devices, using low side switching. If you really need high sided switching, that will be a little more work, and a little more complex, ad a little less efficient, given the same amount of silicon, so if you can get around that all you need is the MCP1630, a MOSFET, a Schottky rectifier, a few Rs and Cs, and of course something to provide the clock.

 

[treez]

Hi

VE1BLL....your idea of simply putting a cap across the LEDs and putting this across the 24V and then dimming by duty cycling a series FET is great...and what we would consider for the future if we do move wholesale into pulsing.

-However, at first, we want to keep our current elec/mech/control fixtures so we can't use the 24V direct, as the LED board (which we don't yet want to change) has 18V across it's LEDs at full brightness and 24V (even pulsed) would overvoltage these LEDs.
-We are keeping things the same as much as possible at first because we still have to repair/replace old stock and service those customers who have already bought our existing Power/Control unit.

Also, big changes are'nt going to wash with the business folk here, who don't see a great need for change whilst our linear regulator LED dimmer-flashers are already very pretty.

jimkirk...I admit a higher frequency than 25KHz would be good for the SMPS, but this product is largely thru-hole assembly and DIP sockets don't like the higher frequencies.
-Thru-hole is easier/safer to repair, especially when at the customer's premises etc.
Your idea of MCP1630 is a revelation to me....it's circuitry is very similar to what we have but would mean less IC's so it's really very good.

macgyvers2000....unfortunately i cannot give my location away as the boss might find out.

 

[MacGyverS2000]

Things just aren't adding up for me on this one. Your requirements continue to seem somewhat arbitrary.

You seem to believe a DIP socket would be a problem over 25kHz... why? If you question the ability of a DIP socket to handle a few hundred kHz, I have to question your ability to continue along the design path of an SMPS. Obviously your design calls for some form of PWM and a more efficient power supply, but you don't yet sound ready to handle a design that will require careful layout of the circuit to avoid harmful oscillations. You also use BJTs rather than more energy-efficient FETs... at hundreds of Watts, these are things you just can't ignore.

I'm not trying to be rude, but it sounds to me like this project (or at least this portion of it) may be more than your current abilities and experience allow for. If you're talking about thousands of units, even the smallest error can result in many recalled units or significantly shortened lifetimes. How happy is your boss going to be if that happens? It may be time to admit to your boss that outside help is the safest and cheapest way to go.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[itsmoked]

I have to agree with Dan a bit as you keep thinking 24V will over voltage your 18V strings. If you PWM the string you don't "over voltage the string", as the PWM is a form of voltage control.

Dips easily handle 100kHz. There are parts running at 20MHz in dip parts.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[treez]

Thank-you, i do appreciate your replies...especially regarding SMPS layout..........

"you don't yet sound ready to handle a design that will require careful layout of the circuit to avoid harmful oscillations"

&

"I'm not trying to be rude, but it sounds to me like this project (or at least this portion of it) may be more than your current abilities and experience allow for. "


....no rudeness detected....your replies are very informative. I will take your point regarding MOSFETs vs BJTs.....regarding PCB layout for SMPS, I doubt that i can go far wrong if i do the following.....

1) Keep power traces as short and wide as practicable.
2) Keep current loops as small as possible. (e.g. having a ground plane return right under the "hot"/signal wire)

 

[GonzaloEE]

PCB layout will help you handle noise/EMI, though the big current spikes from LEDs would negatively affect the remaining loads you just mentioned.

You may add some coil choke between your LED VCC wire and the remaining loads VCC line, to keep current ripple low.