PWM CC

Bartek Banachewicz

Sep 22, 2017 08:50

hmm, I think I'm still not getting the idea of combining a constant current source with a PWM modulation

like, it doesn't seem that simply interrupting the circuit would produce credible current results

PlasmaHH

certainly not the right way to measure wind speed...

Bartek Banachewicz

I'm trying to understand this

PlasmaHH

lots of text, tl;dr but driving an LED with a pwmd constant current sounds like dimming, though I probably would not do it that way but build an adjustable current drive

Bartek Banachewicz

@PlasmaHH what's the difference?

PlasmaHH

@BartekBanachewicz flickering

Bartek Banachewicz

Sep 22, 2017 09:01

@PlasmaHH isn't that how most e.g. automotive LED headlights are driven?

They all flicker in high-speed shots

well, I was specifically aiming for high efficiency, as this is supposed to run off a battery

PlasmaHH

yep, they suck

you know buck converters?

Bartek Banachewicz

@PlasmaHH yep

before you go any further though

so far my plan was to measure the current with a MCU and adjust pwm accordingly controlling a mosfet

I've opened the TL4242 diagram and well

PlasmaHH

many ways to skin a cat

Bartek Banachewicz

Section 7.2

that's just what it is.

PlasmaHH

the key is that since LEDs are current driven what you want is to have the control loop control the current, not the voltage as it is traditionally done

Bartek Banachewicz

Sep 22, 2017 09:05

@PlasmaHH again, the idea was to use a known resistor and measure voltage drop on it

but this can't be done directly since the lower that goes, the lower the voltage you're supposed to measure drops

and obviously the lower the better because it's just causing losses there

PlasmaHH

you can amplify the voltage drop over a shunt resistor

thats how daves µcurrent does it and likely his gw121 too

Bartek Banachewicz

@PlasmaHH what's a shunt resistor?

PlasmaHH

the resistor you measure voltage drop over to get the current through it

Bartek Banachewicz

oh okay

@PlasmaHH oh yeah. The PDF I linked uses an op-amp there if I'm reading this correctly

PlasmaHH

you can also have the µC create a voltage that sets the wanted current for an all analogue boost converter

Bartek Banachewicz

Sep 22, 2017 09:08

that was my idea as well

@PlasmaHH doesn't that operate on voltages as well?

PlasmaHH

lets step back a moment and look at a linear regulator like thing. you have a voltage going into an opamp, and the opamp drives a pass transistor and gets the feedback from the transistors output voltage. so that voltage at the beginning controls the voltage of the transistor

then you can change that into taking the voltage drop over a shunt resistor as an input, so it gets a current source.

Bartek Banachewicz

@PlasmaHH I see

PlasmaHH

then you make it so the reference voltage comes from a µC and you have a linear current source controlled by a µC

Bartek Banachewicz

so the only change here is that the uC doesn't have to do the analog measurement

PlasmaHH

then you replace the opamp+transistor by a buck converter that gets the feedback from the shunt resisotr

it also doesn't control the pwm that drives the buck converter

usually a µC is rather slow compared to opamps and stuff to react to necessary changes when the load or input voltage chaange

Bartek Banachewicz

Sep 22, 2017 09:14

well, in this case the only changes should be the LED warming up and battery draining; I've already constructed the battery measuring circuit which seems to work well, even if it's a tad inaccurate

I think I like the idea of having the current set outside of the mcu though

PlasmaHH

as I said, many ways to skin a cat

Bartek Banachewicz

@PlasmaHH this part I didn't understand though, despite my best efforts

so in the opamp case, the measured current (drop over shunt) is added to the PWM from uC and fed to the transistor

PlasmaHH

nope

Bartek Banachewicz

oh

so I got lost even before that ;)

PlasmaHH

at least not if you mean the case I described

Bartek Banachewicz

Sep 22, 2017 09:17

> you have a voltage going into an opamp

PlasmaHH

you know how the basic job of an opamp is to drive its output in way that the difference between its innputs is 0 ?

Bartek Banachewicz

@PlasmaHH well, I thought I did

PlasmaHH

ok, so let the µC produce a voltage, say 1V

you feed that to the opamp as the reference

then you feed the voltage drop of the shunt resistor to the other side, as the feedback, and the output of the opamp drivers the transistor

so the opamp will now do whatever is necessary to the trasnsitor to make the current through the shunt drop 1V

Bartek Banachewicz

@PlasmaHH I think I sorta meant that but worded myself badly

PlasmaHH

so if the shunt drops 1V at 1A you make the µC output 2V to have 2A curretn etc.

Bartek Banachewicz

Sep 22, 2017 09:21

ok, how about when it drops very little? like 0.001V?

I obviously can't output 2000V from a pin

PlasmaHH

you have two choices (or anything in between that). You create such a low voltage, or you amplify that voltage

2000V would be 2000A ^^

Bartek Banachewicz

oh. right

PlasmaHH

so it also depends on the wanted range.

Bartek Banachewicz

@PlasmaHH ok so I kinda see where I got lost, I mixed two things we're doing here: amplifying the shunt voltage drop and keeping the current constant.

PlasmaHH

0.001V to 10V over a shunt is going to be rather difficult

Bartek Banachewicz

Sep 22, 2017 09:23

@PlasmaHH well the LED is rated at around 3.7V

PlasmaHH

you drive one or many of them?

and at what current range?

and what input voltage? 12V car battery?

Bartek Banachewicz

@PlasmaHH it's one Luxeon K2, supposedly 5W. Driven from a 6V 5-cell NiMH pack

that being said, I am not sure if the lamp itself doesn't have some additional compoents; I don't see the LED, as that's enclosed, and there are 3 wires going out. I am not even sure what the third one is for.

PlasmaHH

what does the datasheet say?

Bartek Banachewicz

@PlasmaHH that in order to get the maximum power it shouldn't need more than 3-something V, but I tried with 5 and only got about 60mA of current

I haven't actually tried bringing this up even more. Now that I have the PSU it should be safe to kinda crank it up a bit

PlasmaHH

you might have fried it, don't put 5V onto a 3.7V LED

Bartek Banachewicz

Sep 22, 2017 09:28

it still lights up

well I was observing the current and slowly raising the voltage

and the current and produced light was nowhere near what I'd expect

PlasmaHH

Have you observed the drop over the LED? also, even when it lights up it might be damaged. Also you might have a fake device, who knows.

and you should cnosult the datasheet about the third "wire".

Bartek Banachewicz

@PlasmaHH I hooked the meter right next to the battery. And it's not fake, it worked before, it's just the controller that died.

PlasmaHH

you should measure over the LED

Bartek Banachewicz

@PlasmaHH There's none.

PlasmaHH

no datasheet?

Bartek Banachewicz

Sep 22, 2017 09:31

@PlasmaHH I mean, the LED is inside a lamp, right. So the LED datasheet doesn't mention the third wire, it's something the manufacturer might have put there

like a temp sensor for example

PlasmaHH

uh, oh, so further disassemble it? ^^

Bartek Banachewicz

@PlasmaHH I'd rather not if I don't have to I guess

PlasmaHH

anways, back to the shunt resistor. Lets take 100mΩ and when you run 1.5A through it that means 150mV drop that should be plenty of room to deal with

you will ahve little choice, obviously its not just the LED in there, or you have a different device than you thought

Bartek Banachewicz

@PlasmaHH I guess

well I'll try taking this apart today then to see what's actually inside

@PlasmaHH so my PWM would need to range from 0 to 150mV?

PlasmaHH

It should be not much of a problem for a µC to create some voltage between 0 and 150mV and for increased accuracy you could use a divider and have the µC create higher ranges

Bartek Banachewicz

Sep 22, 2017 09:34

yeah, I was planning to operate on 3.3V, so 0.15 is about 5%

PlasmaHH

(if you want to use the µC pwm to create that voltage you obviously have to filter it to make it (near) DC, and if the µC has an analogue input pin you can measure that voltage too)

Bartek Banachewicz

wait, measure the voltage I'm putting out myself to control the opamp?

PlasmaHH

0-3V divide by 20 could work

yes, pwm isn't a voltage itself, its just a square wave, you have to filter that to make it like DC ... if your µC has a DAC output then that would be easier though

Bartek Banachewicz

I don't think it does

but yeah, good point; so far I've been driving a MOSFET directly, with the idea that the average needs to be right

in this case it would probably make sense to make sure it's an actual DC voltage

PlasmaHH

so what frequency is your pwm? make a simple low pass filter way below that and you have it output mainly DC

feedback that DC to the µC for increased accuracy and adjust pwm accordinly

Bartek Banachewicz

Sep 22, 2017 09:38

@PlasmaHH I didn't think that would ever be necessary, huh.

@PlasmaHH I was planning to use ATTiny 13

PlasmaHH

well maybe, maybe not, as I said, increased accuracy, many ways to skin the cat

Bartek Banachewicz

way too many :)

PlasmaHH

at least anything like this will make it much easier on the coding side of the µC and it will make it possible for you to split it into two independent parts

you can program the µC to output the DC and until you get that right you can feed into your opamp some DC from another source taht is not the µC to test that and at the end connect both modules

imho much easier as a project when you are not experienced

Bartek Banachewicz

@PlasmaHH yeah I've been trying to modularize it, so far I've split it into 4 parts - battery measurement, UC power, UC itself/PWM, power control

Electrical Engineering

Participants