PIC based buck regulator

[SteveyG]

Hi everyone 

When I finally get my own house at some point this year I plan to run a CAN bus or RS-485 network round every room. The intention being that this will control the ambient lighting in each room which will be RGB LED based (3W, superflux, 5mm etc depending on the fixture). The 3W stars as always pose the problem of the best way to drive them. Assuming I run a 12 or 24V supply to the lights to help prevent so much voltage drop, I'd be wasting as much as 80% of the energy with a linear constant current regulator, so this weekend I quickly threw together some components and some code to test how well I could make a switching power supply.

The main factor here is cost and since most switching regulators are fairly expensive over here in the UK, I wanted to try to make use of the PIC10F series of microcontrollers since they're mega cheap (£0.20 to £0.27!). Here's my first attempt...

(Excuse the awful matrix board - I hate using them, but I couldn't be bothered to go out to the workshop to cut up some FR4 board or develop and etch it  )



Unfortunately I'd somehow ran out of 5V1 zeners for the PIC supply, so that 7805 is supplying the 5V 1mA requirement of the PIC 

Here it is driving a 3W red LED



We get some nice waveforms like this like we were taught at university but didn't believe... 



The switching frequency is a little low, but I only had 220uH inductors to hand. I'll increase the frequency when I get some smaller ones.

I then measured the efficiency of the unit:



80% efficiency isn't too bad for a two hours work  The bipolar transistors are holding this figure back I suspect, although 80% would be perfectly acceptable for me.

Finally all three LEDs driven by the circuit:



When I get chance next weekend I need to check that it works as expected when fed with a PWM signal, if it does then I can concentrate on the CAN bus or RS485 side of things.

Steve.

[justDIY]

Nice work Steve, and welcome to the forums!

So one switcher as a constant current source for the primary colors, and the plan is to dim each color using pwm on the low side of the led?  Won't that cause wild load-transients for the switcher to deal with?

Hows about using a slightly beefier pic (like a 14 pin), and driving three switchers, then you can connect your rs485 directly into the pic, and have it adjust the current output - achieving linear dimming, instead of pwm dimming.

Nice work - I look forward to seeing your project develop.  Any chance you can share a schematic or some source code?

[SteveyG]

No the plan will be to have one regulator per colour, driven by a PIC16F737 which has three hardware PWM channels. If I choose to go the RS-485 route, then the protocol will be DMX-512 style, so it might be difficult to implement three switching regulators on the same device without flickering - even at 10 MIPS. This is something I'll have to try when I've made some further decisions... 

[abnormal]

Hi Steve, welcome!

I don't know how hard set you are on the PIC... the Atmel parts (no I don't work for them  ) may allow a flicker free light as the are true risc. I switched after being PIC for many years and am very pleased. Plus there's avrfreaks.net for help and apps.

Nice project, please let us know more as it develops.

[justDIY]

abnormal, have you any specific examples of a PIC's instruction time being a problem?   I'm not questioning you, just curious to know what the trouble was.

[SteveyG]

Hi abnormal, I'm fairly hard set on PICs because I use them a lot at work and get given free development devices and virtually unlimited components from the guys there (even the expensive ones like the ICE and new RealICE ). We've found that support is also a lot better from the Microchip team so I should really support them...

In terms of the flickering, I can't see how they'd get around the problem since you have to interrupt the switching in order to service an interrupt. This is the same on any device of this level. DMA would be nice - but I don't think I'll find it on a sub 50p device

[abnormal]

multi-reply...

Hi JUSTDIY,

Question away.... 

The PIC has been great don't get me wrong.  Just when it comes to embedded control, higher volume production ($)... and you’re up against the wall "t" state, power and memory space wise, having more registers (not just the "w") and no "clock divide by 4" is pretty nice.  I'm no programmer by any stretch (old hardware guy) and found the Atmel parts have greater throughput and are very easy to use. Having experience with Freescale, Zilog, Philips, Toshiba…  and just about everyone else in the 8 bit uC world they just get my vote at this time.

Hi STEVEYG,

Agree, MC support is great, but I've had issues with other stuff. Not that it would stop me from using them as I still do.
Having the tools is a big big plus, I agree. Just wanted to point out some options that may help.

 - abz

[Rob]

Welcome to the boards SteveyG. You are a bit-tech regular, right? Thanks for dropping by.

[SteveyG]

Thanks for the welcomes 

Abnormal: I'll look into some of the Atmel devices, but I think I would suffer the same problems in any low-end microcontroller. The PIC24's have DMA which is great - but they're a little excessive for each node. I am planning on using a PIC24 for the master controller though 

Rob: I hope to be a regular poster here too! I have lots of projects I'd like to share 

[SteveyG]

Did another few hours work today on this.

After getting the regulator working properly last week, my next task was to make sure I could actually vary the brightness of the LEDs. The best method being to alter the reference voltage that is used to maintain a constant current - so I got a few MCP4921 DACs and interfaced them with a PIC micro. This gave me a massive 12 bit resolution, but at the expense of needing more board space I thought through various other methods, but ideally wanted to keep this as small and simple as possible. Since I plan to use a PIC16F767 for most of the nodes (it has three hardware PWM outputs with 10 bit resolution at up to 19.53kHz) I decided to try a much simpler method than the DACs and just clip (at ~0.6V which is my maximum Vref for 3W LED) then filter the PWM outputs.

Below is the circuit on proto-board again  . Top left is the PIC16F767, top right is the MCP4921. The cream coloured multiturn variable resistor sitting near the ICSP header is connected to an ADC input for testing purposes to alter the brightness. To the right of the variable resistor is the PWM clipping/filter. Below that is the SMPSU section (PIC10F204, transistor, inductor etc.)



A pin header with a jumper allows me to switch between the DAC's output and the filtered PWM output. It turns out that I overlooked one thing when thinking about using a DAC - that the minimum output is -Vref + 1 step, which in this case was just under 1mV which is enough to keep the LED dimly lit  But the filtered PWM output worked beautifully  and eases any board space and layout issues. Yay!

And to end this weeks' update, two pictures of an LED at different brightnesses 





Steve.

[cadstarsucks]

Thanks for the welcomes 

Abnormal: I'll look into some of the Atmel devices, but I think I would suffer the same problems in any low-end microcontroller. The PIC24's have DMA which is great - but they're a little excessive for each node. I am planning on using a PIC24 for the master controller though 

Rob: I hope to be a regular poster here too! I have lots of projects I'd like to share 

All the advantages posted for the PIC can be had for the ARM LM3S101.  The biggest problem with it is it only comes in a SOT28 package.

A quick comparison:
                PIC           AVR               ARM
part         16F636     ATTINY2313  LM3S101
price        2.55          2.24              2.20            (US$)
FLASH      2K             2Kx8             2Kx32 (16 bit instructions)
RAM         128x8       28x8              512x32
MHZ         20            20                  20
USART      no            yes               15550
SPI           no            yes               SSI 16 byte buffers
OSC         8              8                   7.373
PLL          no            no                 yes
analog     2 comp    1 comp          2 comp
mult         no           no                 32x32->64 2-7 cycles
div           no           no                 32/32 2-8 cycles                 
timers     1x8,1x16 1x8,1x16       2x32 or 4x16 or 1x32+2x16
registers 64x8        32x8             16x32

Most hobbyist applications would not need that kind of horsepower but it is the direction the industry is moving and there are 3 different free professional development packages for it (limited by code size), web as well as local cycle accurate simulation, Assembler, compiler (C and C++), linker, RTOS, and hardware debug all free.  (if you are willing to hack up a few cables, if not there is a  $40 dev kit with all that included and a OLED display)

Dan

[thundar]

SteveyG,

I'm trying to build a very similar circuit using a bx-24 (atmel based) uP module.    Are you willing to show the schematic for your circuit? 

[justDIY]

hi thundar, welcome to the boards!