Help designing a DC-DC step-down buck converter

[Nonyaz]

I decided that a step-down buck converter was the best option for my project, I need to drive ~100 LED's from 4 AA's and after giving up trying to cram ~100 resistors and their traces in my board, I figured I'd save board space and be more way more power efficient if I just put them all in parallel and ditched the resistors and gave it the forward voltage directly.

This project is meant to be a christmas present, it will be my first PCB design, and first time using PCB software, and as an added constraint my Altium licence will expire sometime in November so I'd like to have this done by then.  But none of this is directly related to my reason for posting, but since this will be my first post on this forum, I figured a little background was in order.

So I've downloaded a bunch of PDF's on the subject and my first question is about the capacitors.  All the guides clearly want me to use ceramic capacitors for their low ESR, which I didn't even know was a thing, I thought the only ESR I had to worry about was due to the impedance.  So I'm not exactly clear on why caps have ESR, and why ceramics have a lower ESR and electrolytics are higher.  I have a bunch of electrolytic caps of overkill size for this project, but 0 ceramics.  I would like someone to strongly make the case for me to buy a ceramic cap, and if ends up just being to save an ohm then its not worth it to me.

Also regarding caps, only one of the PDF's mentioned in addition to being worried about ESR, I should also fear ESL of capacitors.  This property is more mysterious to me than ESR and I have no itution on the what why and how on this, can someone please enlighten me?  I am a Junior EE major and this is the first time I've heard of ESL in relation to caps...

I'll leave it there for now, thats what's currently hanging me up on this project.

[mariush]

Well, this is an interesting project. 

First of all, let's see what you have to work with... you say 4 AA batteries and 100 leds. 

Assuming you want to drive each led with 5-10mA, you're looking at 500mA to 1A for 100 leds.
A good quality alkaline AA battery will discharge from 1.65v to 1v in about 40 minutes, if you draw 1A from it, so basically with 4AA batteries you'll have about 4v-6.5v for half an hour if you connect them in series.  With less current batteries will last longer, for example about 9 hours if the current draw is about 200mA -  see the graphs at http://www.batteryshowdown.com/

The forward voltage of most leds is 2v to 3.2v so a buck regulator will convert 4-7v to 2-3v with about 70-85% efficiency, so you're looking at under 30 minutes of runtime from 4 regular AA batteries with 1A current draw, maybe 2-4 hours at 3-500mA.

Datasheets recommend ceramic capacitors for low esr but most mention they're preferred to electrolytic capacitors because most CHEAP electrolytic capacitors have high esr. These days there are low esr series of capacitors which have low esr (nichicon hm, hn, panasonic fm, fr, united chemicon kze, ky, rubycon zl*  etc).

I'm not really fond of paralleling leds .. some leds have better specs, some less, they won't have the same current and some may burn up and so on. 

It makes more sense to just use some led drivers.  There are some which can run using a wide range of  input voltages, so you may be able to run the leds directly from battery, and allow you to set the current for each led.

For example, here's a led driver that can control 16 leds (well, 16 channels, nothing stops you from paralleling/serializing several leds on a channel as long and set a high output current), can set the led current from 3mA to 45mA and the input is basically a shift register, so you can just use a microcontroller or something to shift in a few 16 x 1s to light up all leds:

http://uk.farnell.com/texas-instruments/tlc5925idwr/led-driver-constant-current-soic24/dp/1858096RL

this one's similar :

http://uk.farnell.com/texas-instruments/tlc59281dbqr/led-driver-constant-current-ssop/dp/2323439

They kinda work with maximum 5-6v, so maybe you could run the whole thing directly from 4 rechargeable batteries, or you could use a single 4.2-4.6v lithium battery.
 

[Nonyaz]

Well the red LED's are the hogs, they are rated at 20mA, and I need to use all 20 of it to get them at a respectable brightness.  I can underdrive the blue and get a nice pairing around a few microamps, but yea, I'm looking at 1.6A+ of current.  However they will be operating in an outside on, middle on, inside on, all off, repeat pattern so a run time of an hour or more is more than enough.  This is designed to be an on, "aww isn't that cute", off device.  Forward voltage of the red is 1.8v ish.  My entire goal for this project is just to get the PCB made, my first PCB, I don't want to complicate it so much that I never finish it, I don't care if its pretty, I don't care if the EMF from my buck due to my inevitably poor routing deorbits a satellite, I just want it made.

I'll look into a LED driver, however I think I'm committed to this buck converter, I have the possibility to turn this into my free final project for my EE class and so that's a huge win for me there as far as finding time to work on this and school.

I originally wanted to run just 2AA's, but the current kinda killed that, plus all the buck converters I were looking at had a min input voltage of 2.2ish and those AA's loaded would surely dip below that in quick order.  I really wanted to use LiPo's, I have tons from my old RC Heli days, but the ones I'm willing to sacrifice would overhang from the pcb and look ugly.  Plus I already have a AA battery holder, just some quick glue and soldering of the wires on the board and I'm good to go.

Here is my layout now, I haven't deleted the resistors yet, you could see why I gave up trying that approach.  I need to fit 2 buck circuits, 3-4 555's and several transistors on this, want to keep the middle clear for some slik screen.

[Bertho]

This is a design in which you do not want a buck converter. I assume you light up all LEDs together and that you do not have any animation running, although that could still be an option.

The problem with parallel LEDs is that you cannot match the brightness. The brightness of LEDs is current based and the voltage will follow from that. Having two LEDs in parallel means they have equal voltage and that may mean widely different currents.

You want a circuit that enables you to have all LEDs at the same current and that means you need to have them in series. However, 100 LEDs in series requires a very high voltage. You will need to compromise to assure that no lethal voltages are required.

The standard way to drive a long chain of LEDs is to use a step-up converter that is driven with a current feedback. The converter delivers exactly the required voltage to sustain a constant current.

Power requirements will be high, very high. Your design has 94 red LEDs and 45 blue LEDs. With 20mA current and an average voltage drop of 2V over red and 3.2V over blue calculates to 94*2*.02+45*3.2*.02=6.64W as a minimum requirement. Assuming 80% conversion efficiency means 8.3W effective power input (or about 1.4A at 6V). Batteries of type AA have an energy content of ~2500mAh, but that is significantly lower with high loads and the voltage rapidly drops. I fully agree with mariush that your design will last for a very short time on batteries; less than an hour and maybe mere minutes because alkaline batteries release energy only slowly.

BTW, your LEDs seem awfully closely spaced on the PCB. Are you sure you can place them that close physically?

[Nonyaz]

I put a bunch of LED's in parallel on my breadboard already, they are all close enough in brightness, I don't care if a few are 25mA, and a few are 17mA or whatever, this is not designed to have a long lifespan, I have extra LED's and I plan to solder the power circuitry first so I can manually match and throw out any outliers.  Yes they will fit physically, I measured them with my micrometer and made my own footprints, the traces... will fit somehow, you PCB design purists might want to close your eyes for that one

But this is all straying from my original concern, which was about caps, ESR, and ESL.  So far no one has convinced me on the need for ceramics.

[DutchGert]

All C's have a R in them, it's what we call parasitic's

Electrolytics are build way different than Ceramics so there paracitics are a lot different. DC-DC converters basicly rely on these parasitics to become stable, otherwise, an ideal step-up/boost converter would blow itself up.

Same story for inductance, very C has also a small L in them, that means the L in your switcher is not the only L in the circuit. Thats why u have to keep it in mind.

Also, Electrolytics have a polarity and some switching circuits could damage them.

It's not always the bigger capacitance the better at DC-DC outputs

[mariush]

I encourage you to consider using smd leds... you can get some relatively high brightness leds in 1206 footprint ( 3.2mm x 1.6mm) that can do 230-550 mCd at 2v/20mA :

http://uk.farnell.com/kingbright/kpt-3216surck/led-1206-red-230mcd-630nm/dp/2099248
http://uk.farnell.com/kingbright/kptl-3216surck/led-smd-1206-with-optic-red/dp/2217894

and for blue 3.3v / 20mA:

http://uk.farnell.com/kingbright/kptr-3216qbc-d/led-smd-1206-reverse-blue/dp/2217987

The pads will give some extra heat dissipation and at the same time as you will not have so many through holes you can put stuff on the other side. You could just use a hot air gun (or a frying pan) to solder the leds in place after you put them on pads manually.

Those 16 channel led drivers I linked to in a post above can output the current you set at up to about 17-20v (first link chip says max 20v, second link, diff. chip, says 17v) , so you could put let's say 8 red leds in series or 5 blue leds on each channel of those led drivers.
No resistors, straight from batteries, the led driver chips handle the current limiting, you get consistent brightness etc.

From place to place you can just have a via connecting a chain of leds to one of the channels of the led driver that's on the backside, powered directly from the  AA batteries, without buck regulators and all that.  With about 3 such led drivers spread on the pcb backside you can turn on and off all the leds.  You can use a small microcontroller to just shift bits into the led driver chips to turn rows on and off.
The money you save on the buck regulators and the 555 chips can go on can go into the led drivers and the extra money you maybe need for the smd leds.

[Bertho]

I put a bunch of LED's in parallel on my breadboard already, they are all close enough in brightness, I don't care if a few are 25mA, and a few are 17mA or whatever, this is not designed to have a long lifespan, I have extra LED's and I plan to solder the power circuitry first so I can manually match and throw out any outliers.

It is one thing whether or not it somehow works. It is another to have it work correctly. I do not want to be rude, but doing stuff that is known to be a bad idea is simply what it says, a bad idea. Parallel leds is a bad idea and exceptionally poor design. You say you are a "Junior EE major", so you should pay a bit attention to what is right and wrong. Bad habits tend to persist and can lead to catastrophic failure later on. Don't go that way, please.

Yes they will fit physically, I measured them with my micrometer and made my own footprints, the traces... will fit somehow, you PCB design purists might want to close your eyes for that one

Actually, I don't care about the purist view. I just saw the design and I thought "does it fit?". If it does, well then, fine. No need to get agitated about it.

But this is all straying from my original concern, which was about caps, ESR, and ESL.  So far no one has convinced me on the need for ceramics.

ESR: Equivalent Series Resistance
ESL: Equivalent Series Inductance
Both are descriptions of parasitic elements of capacitors. No ideal component exists, so you need to handle the limitations in the design and cope with it. ESR causes a capacitor to dissipate (heat up when current flows). The ESL is a measure how much the capacitor resists changes in current. It should be noted that the physical PCB layout also has influence.
There is no absolute need for ceramics. The point is that the capacitor(s) you need depends on your design and design goals (like stability and efficiency). Some designs can cope with aluminium caps and others really need ceramics and some can go anywhere in between. Often you also have cost considerations. As long as you do not publish a design, there is no possibility to estimate which would suffice for the particular job.

[Codemonkey]

The point the others are trying to make is this:

If you connect your 100 leds in parrallel, and then add a single series resistor to limit the current to say 100 * 20mA, what will actually happen is some LED's will take more, some less. Since a few will be taking a fair bit less, the excess will be taken up by the rest, so much so that some will likely fail, and when they do the same amount of current will now be shared by less and less LED's until they all fail in what I suspect will be a quite spectacular manner. Its a really, really bad idea 

[Marco]

He doesn't want to use any current limiting resistor, he just wants to use a very conservative voltage and just accept they will be relatively dim and perhaps uneven.

[Bertho]

He doesn't want to use any current limiting resistor, he just wants to use a very conservative voltage and just accept they will be relatively dim and perhaps uneven.

The real problem here is that the LED diode voltage drop decreases with increasing temperature. That means when one led gets hotter, then it will start to consume more and more current from the parallel set and burn itself out.

The problem is that a LED gets into a positive feedback loop with temperature at constant voltage, killing the LED at a pace that is faster than I can write "I told you so" 

[Nonyaz]

I get it, its not ideal, its not "right", its not clean, and it makes you guys wince.  But the fact is for what it is its fine, and I highly doubt I will be in a thermal runaway situation, and if it does, its no big deal, its just a silly light-up heart.  I'll learn more from my ideas failing then playing it safe, I need to explore the limits of things.  It would be nice to do the LED controller, maybe put them in pairs with a microcontroller and be able to do some fancy patterns, but the effort to build the supporting stuff to run an aurdrino, and then to program it, its just not in the cards.

I'll post more details as I work on this project, It'll be fun

[Marco]

The problem is that a LED gets into a positive feedback loop with temperature at constant voltage, killing the LED at a pace that is faster than I can write "I told you so" 

The loop gain is smaller than 1.

[Corporate666]

I get it, its not ideal, its not "right", its not clean, and it makes you guys wince.  But the fact is for what it is its fine, and I highly doubt I will be in a thermal runaway situation, and if it does, its no big deal, its just a silly light-up heart.  I'll learn more from my ideas failing then playing it safe, I need to explore the limits of things.  It would be nice to do the LED controller, maybe put them in pairs with a microcontroller and be able to do some fancy patterns, but the effort to build the supporting stuff to run an aurdrino, and then to program it, its just not in the cards.

I'll post more details as I work on this project, It'll be fun

I don't get the point of doing something in a shitty way that will produce shitty results when doing it in the right way is no more difficult?  It's like stubbornly demanding to dig a hole with a rake when a shovel is right there next to the rake.

You say that for what it is, it's fine - but the results will probably be some very dim LED's and some very bright ones.  If uneven illumination is "fine", then why not cut the number of LED's to about 30 and let it be uneven and simpler? 

I would certainly not recommend using this as some kind of school project.  I don't think any teacher would accept this standard of design.

[Nonyaz]

I don't get the point of doing something in a shitty way that will produce shitty results when doing it in the right way is no more difficult? [...] but the results will probably be some very dim LED's and some very bright ones.

It would just absolutely ruin your day if this worked exceedingly well in parallel without doing it your way wouldn't it?

I would certainly not recommend using this as some kind of school project.  I don't think any teacher would accept this standard of design.

And I wouldn't, lighting LED's is not hard nor informative, designing and characterizing a buck converter may be (at least informative).

[Marco]

Do you have to design everything yourself? Or would getting a buck converter evaluation board and just changing the sense divider network be an option?

[Bertho]

It would just absolutely ruin your day if this worked exceedingly well in parallel without doing it your way wouldn't it?

You should do whatever you want.

It seems very clear that you have set your mind that your way is the way to proceed. Who am I or anybody else to tell you what you should do; even though you are arguing up against 30+ years of experience in electronics design.

It has been made clear what your risks are. These are your risks. And a risk means that things may or may not work. If it works today, then it may fail tomorrow or the day after. Anyone's guess what actually happens. That is not a good position to be in with electronics design. You want to assure your design functioning today, tomorrow and also in ten years.

Persistence is a good thing. However, persistence with an obviously flawed construct is foolish. You are in your good right to be stubborn, but don't expect others to respect you for that and I will be signing off of this thread.

[Nonyaz]

I will be signing off of this thread.

I'm right behind you, all I wanted was some information regarding caps in reference to these circuits, not a debate on whether or not LED's in parallel will cause the world to end.  I appreciate the people who did respond to my original question (yourself included).

I may or may not post the final product in a few months for you guys to lambast, but it won't be in this thread.

Take care.

[Paul Price]

While  the temperature coeff. a LED diode tends to decrease the voltage across it with increasing temperature(approx 2.2mV/C), and this tends ti cause that LED to use more of it's share of available current, any increase in current also causes an increase in the voltage drop across this greedy LED, so the effect tends to balance itself out.

While the ideal max. LED current may be approx. 20mA for many types of LEDs, a small fraction of this current may be artistically desirable in a piece of electro-jewelry and the varying brightnesses may tend to increase aesthetic beauty and interest.