hi /diy/ need to drive an led at various levels, can't have

Why can't you have it flash?

If you're using your transistor as a constant current source, both MOSFETs and BJTs will dissipate pretty much the same amount of power.
PWM (flashing) drive has much higher efficiency.
If you want both high efficiency and reasonably low ripple, you need some variant of a switching regulator. There are specialized ICs for driving LEDs that way.

Not an expert on electronics but you can try using a high pass filter on the drain side of the MOSFET. Most of the heating occurs at the low frequency components of the waveform. The high pass filter also doubles as a current limiter.

>>903699

You only really have 3 options for dimming an LED:

1.) PWM
2.) Using an opamp and transistor as linear power supply
3.) A switchmode power supply.

For some reason you can't have it flash (although I can't fathom why just using an usually high PWM frequency like >100kHz wouldn't work), so that's out.

A linear power supply will do what you want, but waste a lot of power. There's no way around the waste, as that's the fundamental method by which they operate.

A switchmode driver would be ideal, but somewhat more costly and complex. There are dedicated switchmode LED drivers all over the place, though, so it shouldn't be too bad.

>>903699
>use a mosfet instead of a transistor

Sure thing brah

>>903699
When they say saturation they don't mean turning the gate on 'too hard' the only whay you can do that is by going past the "Vgs" absolute max rating and popping the damn thing. The "saturation" happens when you try get the MOSFET to pass more current then it can handle (from drain to source) which shouldn't be too much of a problem if you drive the gate properly. If you want to make sure you're not saturation look up the I_D vs V_GS chart which will have a bunch of lines that look like:
______
/
/
then go horizontally across from the amount of current you're looking to pass and make sure that horizontal current line intersects with the line corresponding to the Vgs you're providing, preferably before it plateaus.

What you want to do is turn on the "gate" by applying the maximum gate-drain voltage you can (probably going to just be the voltage of your power supply) without exceeding the Vgs max rating and ATLEAST passing the threshold voltage (Vgs(th)) by a couple volts. The higher the gate-drain voltage the lower the "resistance" and power loss.

If you want some really well written comprehensive info you can try read this: http://www.nxp.com/documents/application_note/AN11158.pdf
but is heavy af so just ask if you need help understanding.

I'll cont about the transistance stuff in a sec

File: IdVgs.png (73KB, 585x540px) Image search: [Google]

73KB, 585x540px

>>903864
Damn the ascii art fuked up, I'll just upload a pic.
The "saturation" of the MOSFET is the plateauing of the curve where the MOSFET starts needing a shiton more Vds (and hence also power loss) to get more current through it.
> is there a decent way to calculate the current through it based on input voltage?
So to figure out the current you find/estimate the Id-Vds curve corresponding to your applied gate voltage on the datasheet chart then that curve will give you the amount of current your MOSFET will be passing. Normally for design you would just turn the MOSFET on high enough that you dont need to worry about saturation then just treat it as a short circuit (at most a resistor with resistance R_DS) when is in the on state. This is also why it is more power efficient since its basically just this tiny equivalent resistor in the on state as opposed to to a transistor which always has a voltage drop, even when saturated (which is a different kind of saturation than saturation of MOSFETs).

So normally when we use a MOSFET in your application (for dimming/power control) we use, as other anon's have said, PWM which is just turning it on and off faster than can be noticed or will effect the application and varying the proportion of time on vs time off there by varying the average power.

Since you said you cant have it flash for whatever reason as other anon said you're stuck with linear or switchmode power. Linear supplies generally use use transistors since they aren't so finicky to control in transistance mode, non of that that Id-Vds crap I explained above to control current/voltage. Its actually pretty doable to setup a simple 3 terminal linear regulator IC to do variable constant current regulation like you need to variably drive an LED, Dave has a really good video here with everything you need to know: https://www.youtube.com/watch?v=CIGjActDeoM&index=1&list=PLBF35875F73B5C9B5

>>903874
Switch modes are a good deal more work to design and a linear power supply can be pretty good if designed properly, even better than linear in some cases (but likely not this). I'd suggest just finding a premade module unless you really need the power supply to be embed. They're also covered on the play list above later on if you want more info.

File: seminar-fabrication-and-characteristics-of-cmos-28-638.jpg (59KB, 638x479px) Image search: [Google]

59KB, 638x479px

>>903699
Most of anons point you in right direction.
there a two sub types of MOSFET's
OLD type FET transistor hard to find in this days. and MOS-FET.
"Metal Oxide Semiconductor" -Flow Effect Transistor .
This transistor is very close similarities to small capacitor. Because Gate and Chanel creating a capacitor. tu turn it on you need charge capacitor , to turn off discharge, of course depends from channel type +/-. And thats it. if charging taking to long then time of switching is longer and you will have transmission resistance from maximum close to minimum open.
You just simply want to charge gate as fastest as possible.
But directly driving Gate from CMOS uP gate isn't the best way, because if there is capacitance and you will increase freq then you current charging & discharging gate will raise to moment when your gate isn't pulling or pushing enough, and thats will not open or close gate and get MOSFET in transitional stage for longer time. Lock at possibility of boosting current by passing it threw some bipolar transistor.
For Higher Voltages operation on high speed you will lock to another sub fraction of MOS transistors named IGBT
Isolated Gate Bipolar Transistor.
But they operating on 1000V and flew amps :) Fly-back generators etc.
For your project if you require some kind of flash protection. I will use PWM generator with active feedback.
Put some current sensing resistor in circuit and measure voltage drop, on it. compare with seated signal and give feedback to PWM to step up or step down duty cycles.

have fun good luck.

>>903699
>need to drive an led at various levels, can't have it flash though. even at high frequency.
no need to worry about that, at high enough frequencies (well, not that high), a pwm to an LED will not look like its turning on and off, and instead will look various levels of dimmed

>>903699

look for a proper mosfet with the required ratings, depending on what you actually want to do. figure out the maximum voltage the system is rated for and figure out how many amps you draw. then pick a mosfet from that category or greater with a low resistance.

you don't want to leave a mosfet in a transition mode.

what you instead want to do is pulse width modulation. (bear with me). saturate it every time.

you said you can't have it flash at high frequencies, but that's a non issue. you pick a frequency depending on what type of capacitor and inductor you can get your hands on.

downstream from the mosfet, hook up a cap, and an inductor to rectify the signal, and hook up the leds downstream from that inductor.

now if you say that you can't even have tiny any ripples, then I got news for you: even the AC-DC converter produces ripples. a DC-DC converter can also produce ripples. what you can do to reduce them is to increase the frequency of your PWM, but you can never make it go away completely (but you can easily get the range them into the mV area).

You can use a digital potentiometer. I think that's what they're called. They're Potentimeters that can be controled digitally by a microcontroller.

>>904183
This.

Why don't you dump all the pwm and transistors and everything else. Just configure an lm317 as a constant current source, then control the current with a digital potentiometer?

If you absolutely can't tolerate flashing, why are you using pwm instead of a constant current regulator anyway?

>>903745
>>903972
i don't actually know, its for a friend and its his requirement the led must be driven linearly

>>903749
i have a few led drivers picked out that are suitable but the cost adds up quickly. i'm just looking at options.

>>903853
indereting i'll have a think about this

>>903855
i'm attempting a combination of 1 and 2, 3 is an alternate solution that i appreciate

>>903859
ok.

>>903874
>>903935
>>904027
thanks

>>904183
i noticed these recently, i'll come back to them once i have this method down thanks.

>>904185
i'm using an opamp as a current source but controlling it with pwm through a low pass filter. never thought about an actual regulator though, might be a good plan.


anyway thanks guys you have been great as always.

>>904264
>i don't actually know, its for a friend and its his requirement the led must be driven linearly
To be fair I hate it when you can see the strobing effect every time you move your eye.

>>904343
if it flashes fast enough there's no way your eyes could tell the difference.

>>903853
A high pass filter would not have the desired effect. You want to switch at a high frequency, then filter the switching frequency away, just to have the average signal of the switching signal..

>>903855
A constanc current LED driver for a 1A 3.7v led is not that expensive, just look at eBay, it is like a couple of $.

>>903935
You have gotten some basic things completely wrong.

As to mosfets, there are many subtypes, but an IGBT is not one.
- You have Depletion and Enhancement mode, this refer to the gate turn on threshold voltage.
- Then you can further divide by N and P channel devices.
- Sub device by Lateral and Vertical structures
- Then gating topology.
Today most mosfets are N type Enhancement mode mosfets.

An IGBT is a completely different structure, and only has some similarities with a mosfet. Depending on size, they can handle up to 5.6kV and some 3-5kA..
IGBTs are no where near fast components, they suffer from trailing currents and minority carriers being trapped in the channel when turning off.
- If you want high speed switching at ~800V and above either cascode mosfets or use SiC mosfet for up to 1.7kV

at higher voltages, high speed switching is not really a choice.

>>904264
>>904343
As >>904379 said, switching "fast" will ensure that you are not able to see it. 50Hz should be near the limit, 100Hz is definitely above the visible limit, but 600Hz is no where near fast switching for an LED..

I have had a regular star mounted 1W UV led switch at 6 MHz. It was definitely not build for the purpose, but it worked. If you are a 100 sure you do not want it to flicker, then use a constant curret buck, they can be found on ebay for a couple of bucks..