Picking a transistor.

[paulca]

I only have NE2222 and BC548's, but I have a project that needs more current than they can handle.

Basically switching a 12V line using a 3.3V PWM pin on an MCU.  It's driving some 5050 LEDs, 12V with groups of 3 in series.  Power is meant to be 14.4W/m, so call that 15W/m, I have 2 meters, so that's 30W or 10W per colour, so 10W / 12V ~= 1 Amp per transistor.

Would these do:
https://uk.rs-online.com/web/p/bjt-bipolar-transistors/7390347/

BTW, these are the LEDs with GND, RED, GREEN, BLUE connections, not the digital WS8212s

[MagicSmoker]

That's a PNP BJT optimized for audio amplifiers, so probably not the correct polarity and otherwise not ideal.

Can you post the schematic you are working off of?

[paulca]

Basically this.
https://randomnerdtutorials.com/10-diy-wifi-rgb-led-mood-light-with-esp8266-step-by-step/

But with 2 x 1 meter of strips of LEDs.

A thought did occur to me.  These are to light a fish tank that contains plants.  I could need it to power 4 strips.  60W, 20W per colour.

[ledtester]

The typical solution for low side switching with a microcontroller is a logic-level N-channel MOSFET.

Get one that will fully turn on at 3.3 volts. Searching for "arduino 3.3v logic level mosfet " should bring up plenty of discussions and suggestions.
 

[David Hess]

1 amp will require say 20 milliamps of drive current at a forced beta of 50 for a bipolar transistor which is marginal for most logic outputs.  That means either using some sort of driver to provide higher base current, a Darlington bipolar transistor, or a MOSFET.

So bipolar transistors could be used but the higher cost of a power MOSFET is probably worthwhile for simplicity.  Or a higher gain higher current bipolar like the KSD5041 could mange a forced beta of 100 for 10 milliamps of drive current.  For suitable bipolar transistors, look for higher current devices so that they maintain their gain at higher currents.

[paulca]

Mosfet specs always confuse me.  A few google searches pointed to Digikey listings of "logic level mosfets", but even when I inspected a few datasheets they had figures like Vgs min 2V, Vgs max 4V, but at silly low current.  The RDS(on) voltage was 5V.  So a 3.3V MCU pin might not be able to turn the mosfet on fully.

Of course I don't shop at Digikey so I tried searching RSComponents and they don't even allow you to search by Vgs, but only list Min and Max gate source voltage.  Then some of the datasheets almost look like they are trying to obscure that figure, listing it in strange place with strange names and never just Vgs(on).

[paulca]

To show how bad I suck at finding these things I searched for Mosfets with a Vgs max of 3V. 

I found this:
https://docs-emea.rs-online.com/webdocs/0791/0900766b80791120.pdf

While I'd have to use an adapter board as this project will be proto boarded, I read the datasheet anyway and ... sure enough to get it's RDS(on) figure at 3Amps, do you need 3V... of course you don't you need 10V!  Grrr.

3V will get you 0.250mA.

What am I missing when search for these?  How do I ask the basic question, "Find me mosfets capable of conducting fully at X amps with 3V on the gate?"

... and yes it will be low-side switching, so the voltage across it will be quite low and the (source?) will be at 0V.

[Mr. Scram]

Mosfet specs always confuse me.  A few google searches pointed to Digikey listings of "logic level mosfets", but even when I inspected a few datasheets they had figures like Vgs min 2V, Vgs max 4V, but at silly low current.  The RDS(on) voltage was 5V.  So a 3.3V MCU pin might not be able to turn the mosfet on fully.

Of course I don't shop at Digikey so I tried searching RSComponents and they don't even allow you to search by Vgs, but only list Min and Max gate source voltage.  Then some of the datasheets almost look like they are trying to obscure that figure, listing it in strange place with strange names and never just Vgs(on).

Silly filtering options seem to be a recurring frustration.

[paulca]

So I found this...

It has 95 milliohm resistance for 3A at 2.5V.

https://docs-emea.rs-online.com/webdocs/1332/0900766b8133282e.pdf

But I had to pick a very low Vgs to get that.  Sounds suspicious.  ... and it's SMD only.  I'd have to dead bug these guys.

[paulca]

I realise this would be complicating things, but could I use a jelly bean NPN bi-polar to switch the 12V onto the power mosfet gate?  This would open up the mosfet search and include things that require 10V on the gate.

[hamster_nz]

I've used the now obsolete si4944dy and sy4946dy for switching a few amps with 5V logic.

Si9925DY seem to be is the 3V equivilent for switching a couple of amps.

I was using an optoisolated mosfet driver to isolate the micro, so haven't look if a simple micro pin would turn it on fast enough if you intend to use PWM

[magic]

Zetex (now Diodes Inc) makes good BJTs for saturated switching. ZTX851 might fit the bill with β>100 at 1A, 200 typical. It also coincidentally has some of the lowest voltage noise available in any BJT, supposedly at 200pV/rtHz, but that's another story.

Also check if they make something in SMD, may be cheaper if SMD is okay with you.

Worse saturation voltage but more beta and much cheaper: 2SD965, but I can only find it at TME and Reichelt and some small shops in Poland. Perhaps it's out of production now

edit
CDIL still has CD965 on their website, but good luck finding distributors of CDIL in the West.

[MagicSmoker]

Basically this.
https://randomnerdtutorials.com/10-diy-wifi-rgb-led-mood-light-with-esp8266-step-by-step/

But with 2 x 1 meter of strips of LEDs.

A thought did occur to me.  These are to light a fish tank that contains plants.  I could need it to power 4 strips.  60W, 20W per colour.

Skipping the video and scrolling down the page to the schematic, it shows the expected tiny small-signal NPN BJTs being driven directly by the MCU, but that ain't gonna cut it if you need to switch 20W at 12V or so.

The easiest solutions are either use an n-channel MOSFET driven by a MOSFET gate driver IC, or a "smart" low-side switch IC (which will also likely be protected against overvoltage and shorts, so even better).

An example of a smart low-side switch is: https://uk.rs-online.com/web/p/mosfet-drivers/1630759/

A Zetex (nee Diodes, Inc.) BJT as suggested by @magic above might just work, but it would be marginal, in my opinion. You really don't want to demand more than about 5-10mA per output pin on an MCU.

[electrode]

I like the IRLB8721PbF. Its on resistance is around 10 mΩ @ 5 V (Vgs). However, the datasheet doesn't specify RDson at 3.3 V.

You could get a couple and run some measurements on them – at only 1 A of current, they're likely to handle your requirements without requiring MOSFET drivers (bleh), but no guarantees...

[MagicSmoker]

I like the IRLB8721PbF. Its on resistance is around 10 mΩ @ 5 V (Vgs). However, the datasheet doesn't specify RDson at 3.3 V.

There's always a graph of Id vs. Vgs (ie - transconductance), and from that you can extrapolate Rds[on]. In this case, Fig. 3 shows Id is around 7A at around 3.3V Vgs and with Vds = 15V, so Rds[on] is approximately 2Ω...

That said, upon actually going through the datasheet of the the low-side switch I suggested above it really appears to be nothing more than a MOSFET with a bit of protection circuitry thrown in, so it has a similar transconductance curve... A proper smart low- (or high-) side switch would still be the easiest to implement, if not the least expensive, option. Generally for beginners I err on the side of easier rather than cheaper.

[magic]

The KSD5401 posted by David Hess seems good too.

For low threshold MOSFETs, simply look for SOT23 parts. I went to TME, punched in =20Vds, >5Ids, SOT23 and the first 3 parts it spat out were all rated <30mΩ at 2.5Vgs.

[xlnx]

I have made an attempt to finding modern low-cost transistors that I can standardize on for all my designs. I have found a few which lets me to low-side and high-side switching at 1.8V, 2.5V, 3.3V and 5V. They are:

Name   Pkg           Type  US$    VDS(V)  ID(A)    RDS
AO3400   SOT-23   N      0.01    30V       5.8A    <52mΩ @ Vgs=2.5V
AO3401   SOT-23   P      0.01    30V        4A      <85mΩ @ Vgs=2.5V
AO3415   SOT-23   P      0.03    20V        4A      <65mΩ @ Vgs=1.8V
AO3416   SOT-23   N      0.04    20V       6.5A    <34mΩ @ Vgs=1.8V
AO3407   SOT-23   P      0.02    30V       4.1A    <87mΩ @ Vgs=4.5V

These can be found on AliX for almost nothing, and they have absolutely fantastic specifications.
Data sheets here: http://www.aosmd.com/

[Mr. Scram]

Skipping the video and scrolling down the page to the schematic, it shows the expected tiny small-signal NPN BJTs being driven directly by the MCU, but that ain't gonna cut it if you need to switch 20W at 12V or so.

The easiest solutions are either use an n-channel MOSFET driven by a MOSFET gate driver IC, or a "smart" low-side switch IC (which will also likely be protected against overvoltage and shorts, so even better).

An example of a smart low-side switch is: https://uk.rs-online.com/web/p/mosfet-drivers/1630759/

A Zetex (nee Diodes, Inc.) BJT as suggested by @magic above might just work, but it would be marginal, in my opinion. You really don't want to demand more than about 5-10mA per output pin on an MCU.

These switch ICs can do the job by themselves?

[MagicSmoker]

...
An example of a smart low-side switch is: https://uk.rs-online.com/web/p/mosfet-drivers/1630759/
...

These switch ICs can do the job by themselves?

Not the specific one I suggested above, but, yes - these smart high and low side switches are specifically made for driving loads directly from an MCU output. However, the MOSFETs that @xlnx found above look like good choices, too.

[David Hess]

Zetex (now Diodes Inc) makes good BJTs for saturated switching. ZTX851 might fit the bill with β>100 at 1A, 200 typical.

The Zetex parts should be ideal for this.  Lots of companies make "Super E-Line" transistors now.  The KSD5401 that I mentioned is one of them.

It also coincidentally has some of the lowest voltage noise available in any BJT, supposedly at 200pV/rtHz, but that's another story.

The way I remember it, Zetex bought old and depricated CMOS IC fabrication equipment and used it to make modernized RETs (Ring Emitter Transistors).  You might be familiar with these in the form of the old D44/D45 TO-220 power transistors.  Their construction includes multiple base connections and results is less Ft and hfe droop at high currents.  They are very popular now for linear power amplifier but also work for fast linear regulators and switching regulators.

The multiple base connections should also result in low base spreading resistance which otherwise limits low noise performance but this is not something I have investigated yet.  Common small signal transistors are like 40 ohms but integrated transistors which use similar construction can be 2 ohms or less.

[magic]

They have an appnote which shows the internal construction and lists some part numbers with even more beta.
https://www.diodes.com/assets/App-Note-Files/zetex/an21.pdf

Base spreading resistance is clearly a factor given that their broadband noise is less than 10Ω resistor's. OTOH, dedicated low noise types still beat them at 1/f. Somebody posted a bunch of measurements here:
http://www.dicks-website.eu/low_noise_amp_part3/part3.html

[edavid]

The Zetex parts should be ideal for this.

Except that Diodes has jacked up the price - $1 each for a TO-92 part

Lots of companies make "Super E-Line" transistors now.  The KSD5401 that I mentioned is one of them.

KSD5041 is more reasonable, but you could still find TO-202 or TO-220 parts for less.

P.S. The trouble with some of the beefier MOSFETs mentioned is that OP said he wants to use PWM, so gate capacitance is an issue.

[David Hess]

The Zetex parts should be ideal for this.

Except that Diodes has jacked up the price - $1 each for a TO-92 part

Counting inflation that is still less expensive than they used to be.  But they have always been more expensive than jelly bean parts like the 2N4401.

Lots of companies make "Super E-Line" transistors now.  The KSD5401 that I mentioned is one of them.

KSD5041 is more reasonable, but you could still find TO-202 or TO-220 parts for less.

Or a TO-126/TO-225 part but these packages are not used for newer bipolar transistors so tend to have lower hfe.  For a TO-220, the D44H11 might work with a forced beta of 100 so 10 milliamps of drive current.

P.S. The trouble with some of the beefier MOSFETs mentioned is that OP said he wants to use PWM, so gate capacitance is an issue.

And if you add a gate driver for the capacitance, the advantage of simplicity is lost.

[paulca]

So I tested 1 meter of the LED strips on the brightest white with the silly controller they came with and the whole set pulled 1.2A, 14W.  So that's only 400mA per colour, I think the green pulled 420mA the rest 390mA.

So rather than running 2x1meter strips with the option of going to 4x1meter strips from one set of transistors, I could always create a little array of small transistors.  For 4 strips it would be 12 transistors, but each would only need to handle 500mA.

As a bonus I could control each strip individually.  Although that would require 12 PWM channels, which I would need to check.  As it's for a fish tank I don't need much in the way of control, but I do need a way to shift the colour a little bit, probably dropping the green light a bit to favour the plants preferred spectrum, or favouring the blue to make the tank look prettier.  Then again fading out rather than cutting hard would be better for the fish too, which requires PWM on all colour channels.

Without 12 PWM pins I could of course just run 2 transistors in parallel off one PWM pin to run 2 strings of LEDs from 3 PWM pins and 6 transistors.  Meaning I only need 6 PWM pins which I think is available.

[MagicSmoker]

...
P.S. The trouble with some of the beefier MOSFETs mentioned is that OP said he wants to use PWM, so gate capacitance is an issue.

The OP didn't specifically say the LEDs would be PWM'ed, but it's a fair (even likely) assumption. If the OP does intend to PWM the LEDs then there really is no avoiding either using a MOSFET driver or a smart low/high side switch here. You simply can't PWM any MOSFET bigger than a 2n7000 at more than a few hundred Hz with the typical 10mA available from an MCU I/O pin; not without said MOSFET spending most of the time in its linear region, anyway.

[paulca]

So the options that seem to present themselves are:

* to use small NPN transistors in parallel, so each one is only switching <500mA.  It would mean 12 transistors for 4 strips of RGB.
* bin these silly LEDs and just buy the 12V WS2812s, even though I don't need individual (or groups of 3) addressing, they are much easier to deal with.

I might use these elsewhere and just wire 12V to all the pins (or actually 12V to one and ground to the others).

[ledtester]

This Adafruit page suggests using a TIP120 or IRLB8271:

https://learn.adafruit.com/rgb-led-strips/usage

Here's the datasheet for the IRLB8721:

https://cdn-shop.adafruit.com/datasheets/irlb8721pbf.pdf

Looks like it should be able to do at least 1A perhaps even 2A at 3.3V.