Choosing a transistor

[The Master]

Hi, How do you go about finding the right transistor for the job? Is there somewhere that lets you search based on certain values like the parametric search at micrchip.com?

[skimask]

One quick thought...I usually go to www.digikey.com, whether I end up buying from them or not is another story...
Type 'transistor' in the search box, pick the type of transistor (example Transistor BJT Single under Discrete Semiconductor Products), then narrow it down based on the lists provided. Do it for MOSFETs, chips, resistors, whatever.
I'm sure Digikey isn't the only company that has a 'drill-down' type system for figuring out what a person wants to buy. And I'd be willing to bet that if you know a semi-conductor manufacturer (i.e. Fairchild, Motorola, etc), you'd probably be able to do the same thing at their websites.

[The Master]

I use Rapid for components. They do have a kind of drill down to find things but its not got anything like that site you linked me to. I did a quick search but i cant seem to find any transistors that rapid sell. Maybe im looking for the wrong things though. Now i can find the formulas i need and put the results into that page and hopefully find what i need and hopefully i wont end up using NPN transistors again where i need PNPs

[Squibcakes]

If your desiging a circuit from scratch - then chosing the right transistor is a bit of a black art.

That said, if you have an idea of the operating voltage and current draw you can narrow your search down.

Probably best to trawl the net for some circuits (similar to your idea) and copy that.

If your replacing a stuffed transistor - change it with the same or get your hands on a transistor comparison tables book to find an equiv.

Squib

[skimask]

If your desiging a circuit from scratch - then chosing the right transistor is a bit of a black art.
No kidding! Sometimes it's black enough that you end up with nothing but that... Black art all over your circuit board! :D

[Lajko]

I think a question is what you want to do with the transistor. If for driving something, there are tons of IC drivers for driving high or low and very high current DC or AC loads - even opto-isolated drivers. There are even opto isolated triacs (I've used them to directly drive up to 25 watts of lights for disco effects! 64 channnels for $1 a channel!)
So fill us in and maybe everyone can help with a driver IC or how to select your transistor.

[The Master]

This is actually for disco lights. Im making some that use some full color LEDs. I checked the datasheet of the LEDs and it appears the Red, Green and Blue use 30mA, 25mA and 30mA respectively. The datasheet for my PIC says each IO pin can only take 25mA. I will be using 17 of these LEDs on each light and they have to be wired in parallel so thats way too much current for the PIC to handle. I decided to add some transistors in and they seem to work although im using NPNs at the moment when i should be using PNPs. About 2.5V seems to get the LEDs to the correct brightness without overheating but to reduce the overall current needed im flashing them very fast between Red, Green and 2 Blues so each color is on for 1/4 of the time. Someone on here told me that means i can use 4 times the voltage because it averages out to what it was origionally. I tried that and it seems to work fine.

I would like to use transistors rather than a chip because the PCB is going to be extremely small and transistors can be dotted about where they are needed (2 on each side). I definately have room for 4 transistors but i cant make the PCB any bigger. To make things even more interesting its a round PCB!

Im pretty sure i would want a darlington pair type of transistor. Thats what im using at the moment for other projects and they do the job pretty well. There are 2 problems with them though. The first is that they are NPN and the second is that Rapid have suddenly stopped selling them. I would like to use NPN transistors but the LEDs i have are common cathode

[The Master]

Ive been trying some more values in the search at digikey and its started to give me no results for what i want. I had a quick look through the transistors at rapid and found one that looks like it might work. 2N3702 40V TO92 PNP GEN PURPOSE TRANS RC (http://www.rapidonline.com/netalogue/specs/81-0270.pdf). The most current the LEDs would draw should be 510mA. The datasheet for this transistor says "Collector Current - Continuous 500 mA". It wont be used continuously. It will only be on for 1/4 of the time at the most. It also says "Collector-Emitter Voltage 25 V". I was planning to use 12V for the LEDs (depending on further tests). I am using 12V at the moment but it might be because im using NPN transistors. When i put the correct PNPs in then 12V might be too much for them. The datasheet doesnt mention darlington pair so i assume its just a normal transistor. I think that means that the output might be slightly lower than if it were a darlington pair. That might be a good thing if the 12V is too much.

Edit: That one is also in a TO92 package. It doesnt matter too much but those would fit best on the PCB. Things like the TO220s that ive seen are way too tall. There is also a height limit on the PCB of about 8mm

[Luciano]

Hi,

Use a power MOSFET with logic-level Gate Drive.
(You can drive it directly with the I/O of the microcontroller).

For my Stereo microscope ring light (warm white LED's) I use
IRLR/U2905 mosfets with I-Pak TO-251AA case. The microcontroller
and the LED's use the same power supply of 3.25V.
(20mA @ 3.25V for the used LED). The LED's are driven with PWM.

IRLR/U2905 datasheet:
http://www.irf.com/product-info/datasheets/data/irlr2905.pdf

Best regards,

Luciano

[The Master]

(You can drive it directly with the I/O of the microcontroller).

Should i not be doing that with normal transistors?

I cant find that type of mosfet on rapid's site but they do have a mosfet category (http://www.rapidonline.com/Electronic-Components/Discrete-Semiconductors/MOSFETs).

Would i be right in saying the following
Drain-source voltage is the max voltage it can output to the load
Drain current (continuous) is the max amps it can output to the load
Drain current (pulsed) same as above but relevant when using PWM.

Im not so sure about the Drain-gate and Gate-source voltages. Looking at a few datasheets i would guess that the Gate-source voltage is the max voltage that the PIC chip can output to it. Does it matter if its a lot lower? One datasheet im looking at says +-18V. Does that need 18V to make the LED as bright as possible or will about 3V do the same job?

I know how many volts and amps the circuit will be using so do i just have to look at Drain-source voltage and Drain current?

Im planning a slightly bigger light too so i will need to know things like this. The bigger one will use seperate LEDs for each color so i can get the really bright ones. That means i can wire some in series to use less current too.

Edit: I see there are N-type and P-type mosfets. Do i need a P-type?

[Luciano]

Hi,

Use a power MOSFET with logic-level gate drive.

This is what I use for my lamp:
http://img525.imageshack.us/img525/6549/led1dc2.png

If you use a 12V supply use that:
http://img401.imageshack.us/img401/3978/led2ij5.png
Best regards,

Luciano

EDIT: For the 12V schematic you can also connect groups of the 4 or 5 LED's in series.

[The Master]

Ive found some logic level ones in rapid. They all seem to be n-type though. Does that matter? The circuit im making has to have the mosfet and LEDs the oposite way round. The multicolor LEDs have a common ground so its the positive side that needs to be switched

[The Master]

Would it be better if i redesigned this small light to use seperate LEDs for each color? I cant add anymore LEDs because of space so it would be like dividing the amount of LEDs by 4. I could also connect them in series etc and group their anodes together so i can use either an NPN transistor or the same type of mosfet that you showed in your diagrams.

The only problem with using seperate LEDs is the color mixing. The idea is that they combine to make loads of colors. When the light is shining at an object then its not so bad but these small ones are designed to be pointed at people. I think from a distance the colors should mix alright

[The Master]

Ive been playing around with the circuit design and using single color LEDs ive managed to fit 27 more of them on the board! Im not quite sure how i did it either :S I think this way might work better. I can use brighter LEDs now and instead of 2 blues i can use 1 blue and use the 4th chanel for white. The price of the light has about doubled by using more LEDs but it should be worth it. Is 35,000mcd bright? It sounds like quite a high number.

Can i just double check that a 4V LED will be fine running on 12V if i use PWM? The average would end up being 3V. Im sure it was someone from here who said that would work but i dont remember who.

The alternative is to wire the LEDs in series-parallel and not use PWM when they should be at full brightness. The only problem i see here is that the light would use 4 times a smuch current because all 4 sets of LEDs would be on at the same time

[skimask]

Is 35,000mcd bright? It sounds like quite a high number.
Depends on the viewing angle.
If you've got a 35kmcd LED with a viewing angle of 170 degrees, then, yes, that's a seriously bright LED. If that same 35kmcd LED only has a viewing angle (1/2 theta) of 10 degrees, then it won't be so bright.


Can i just double check that a 4V LED will be fine running on 12V if i use PWM? The average would end up being 3V. Im sure it was someone from here who said that would work but i dont remember who.
LEDs don't run on voltage...get that idea out of your head. Yes, they need voltage...'nuff said.
It's CURRENT that makes an LED work. Yes, you have to have enough voltage to get the current required, which is why you also sometimes need a current limiting resistor. You start pumping 12v thru a single LED (unless it's rated for 12v of course), and you WILL burn it up sooner rather than later. If you've got 12v PWM running at 25% duty cycle, yes, the average would be about 3v-ish, but you're still hitting the LED with 12v. Again, if it's rated for that 12v, then I suppose you'll be fine. Otherwise, look to have a bunch of open LEDs eventually.
Do yourself a favor and get a copy of 'The Art of Electronics' and do a bit of reading. It's my 'bible', great book, will teach you a lot about a lot of stuff.

[Luciano]

Hi,

The choice of using 17 LED's is not going to help you
if you connect groups of LED's in series.

* * *

With the PWM you can control the brightness of the LED's.
With or without PWM the maximum forward voltage is the same.
The maximum forward voltage is visible in the datasheet of the LED.

Best regards,

Luciano

[The Master]

I had a look for that book. £48.45 on Amazon! It wants to be good for that price. Ill add it to my x-mas list ;)

The LEDs im looking at are 15 degrees i guess that puts them towards the dimmer end.

So, using LEDs on 12V in parallel is definately out of the question. Looks like ill be using a series-parallel type circuit. That way each of the 4 sets of LEDs should only use 200mA of current.

If i only need 200mA now is it still better to use a mosfet instead of an NPN transistor? Yes i can use NPN now with these other LEDs

[The Master]

Luciano, I have seen the max forward voltage. I thought it meant on average.

Im using more LEDs now. Ive re-arranged the circuit a little so that theres hardly even room for the 2 screws and ive got 48 LEDs on it. Its a bit less cramped on the track side because each LED only has 2 legs instead of 6. There are 12 LEDs in each color. The datasheet says 4.2V max for the white ones. If i have groups of them in parrallel (4 LEDs in each group) and connect the groups in series then i should end up with 4V per LED. I think that means they would only use 150mA per color too

[skimask]

I had a look for that book. £48.45 on Amazon! It wants to be good for that price. Ill add it to my x-mas list ;)
Obviously, not everything in that book applies, but the first bunch of chapters (i.e. 1/3 of the book) breaks down the basics of transistors/mosfets, and the bulk of the discrete components needed to put together a complete circuit. And it really helps to have a few examples of each laying around, along with some spare parts and a meter (and/or 'scope) to play with pieces/parts and see what actually happens in practice vs. just reading all about it.


The LEDs im looking at are 15 degrees i guess that puts them towards the dimmer end.
Not necessarily. Look at the datasheets for those LEDs, look at the distribution, then compare that LED with the cheaper and more expensive types.


If i only need 200mA now is it still better to use a mosfet instead of an NPN transistor? Yes i can use NPN now with these other LEDs
I think you're better off using a MOSFET over a BJT anyday. The 'bible' mentioned above shows why and I don't remember exactly why. But it's got something to do with the forward bias voltage. A BJT gets hot, forward bias voltage drops a smidge, current flow rises, BJT gets hotter, forward bias voltage drops a smidge, current flow rises, it's a nasty circle of death for the BJT. With a MOSFET, it effectively chokes itself when it gets hot. I don't remember the exact mechanism by which this happens, but it supposedly can 'save itself' in certain situation whereas a BJT would go up in smoke.

[mackrackit]

You might find this useful
http://otherpower.com/otherpower_lighting_leds.html

[Luciano]

Hi,

The power dissipated in a MOSFET is the voltage across it times
the current going through it.

The voltage across the MOSFET when it is on will be the
resistance of the MOSFET, Rds(on) times the current going
thorough it.

A MOSFET IRLR/U2905 has a on-resistance of 0.027Ω.

Power dissipated by the MOSFET case when you have 2 Amps of
current going thorough it will be: (2A = 100 LED's 20mA).

P = 2A^2 * 0.027Ω = 0.108 Watts.

(No problem without heatsink).

Best regards,

Luciano

[The Master]

Yes that is usefull. I think im using the type of regulators they mentioned (the bad ones). The plan was to use a 12V one for the LEDs and a 5V one for the rest of the circuit. It will help picking the correct resistors for the different colors too (each color uses different amounts of amps and volts)

The lights im making are actually like spotlights so i would have thought a 15 degree angle should be fine. There is a directivity diagram in the datasheet (http://www.rapidonline.com/netalogue/specs/55-2484.pdf) Im not sure if thats what you mean. I can try and compare it with some other LEDs i already have but the best thing is probably to buy a few and turn them on

I do have a multimeter and now a PC oscilloscope. I had to use them a lot while getting a test circuit working with the origional multicolor LEDs

I see what your saying about the transistors. I did have some running with the old LEDs though and they seemed to be fine. They did start getting a little hot so i turned my PSU down a little. They cooled back down and continued to run for about 2 hours with no noticable temperature increase. I will order some mosfets instead though just to be on the safe side.

I think what im going to do is try to figure out exactly what mosfets and resistors i need etc for those LEDs then buy a few and see how they work

[The Master]

(No problem without heatsink).


Thats good news then. I dont even have room for TO220 packages. Theres no way i could get a heatsink to fit in there. Im not even using a heatsink for my voltage regulator since it never gets remotely warm with just a PIC chip and a MAX485 connected.

One thing i dont understand is how you know that 0.108W means you dont need a heatsink

[Luciano]

One thing i dont understand is how you know that 0.108W means you dont need a heatsink

Hi,

From the datasheet. The thermal resistance junction-to-ambient is for when
there is no heatsink.

IRLR/U2905 datasheet:
http://www.irf.com/product-info/datasheets/data/irlr2905.pdf

The thermal resistance Junction-to-Ambient of an IRLR2905 is 110 °C/W.
Its operating junction temperature range is -55 to + 175 °C.



Example ambient air temperature = 45°C:

175°C − 45°C
__________ = 1.18 Watts
110°C/W

With an ambient air temperature of 45°C, without
heatsink an IRLR2905 can dissipate 1.18 Watts. (Absolute Maximum).


Best regards,

Luciano

[skimask]

The voltage across the MOSFET when it is on will be the
resistance of the MOSFET, Rds(on) times the current going
thorough it.
A MOSFET IRLR/U2905 has a on-resistance of 0.027Ω.

Of course that also assumes you are driving the MOSFET 'hard enough' (i.e. large enough Vgs vs. Vds) to get that low Rds(on).
If you come up a bit short on the Vgs, the Rds(on) could be much higher.

It's always something, ain't it? :)

[Luciano]

Of course that also assumes you are driving the MOSFET 'hard enough' (i.e. large enough Vgs vs. Vds) to get that low Rds(on).
If you come up a bit short on the Vgs, the Rds(on) could be much higher.

It's always something, ain't it? :)

The power MOSFET IRLR/U2905 has logic-level gate drive and
therefore you don’t have to worry about that. Just try and
measure the voltage drop.

[The Master]

Ive been looking at the datasheets again and as far as i can see the LEDs in total will use at most 200mA per color. Ive been trying to find a suitable mosfet to control them but im still having a bit of trouble. There is a category on rapid called "TO-220 Logic level power MOSFETs N-Channel". I dont want a TO220 package but i thought i would look at the datasheets anyway because they seem to be exactly what you recommended. I still cant seem to understand this gate voltage. Even the logic level one im looking at says "Gate-source voltage ±18V". There is also a "Gate Threshold Voltage" That says 2.5V. Is that what i should be looking at?

I almost got the heatsink bit. skimask confused me a bit though. I have a rough idea of how hot certain components will get with a certain load on them but usually i just put my finger on it to test. That seems to work well except for when i shorted a regulator out once... ouch :P

So, It looks like the only logic level mosfets that rapid do are in a TO220 package. I want to look through the rest of the mosfets to see if any will do the job but it looks like most of them are for audio circuits

Edit: Would the 2N7000 N CHANNEL MOSFET (http://www.rapidonline.com/netalogue/specs/47-0180.pdf) be suitable?

[Luciano]

Hi,

Use that:
http://img440.imageshack.us/img440/8364/led5nc5.png

Datasheet MPSA14 DARLINGTON TRANS NPN:
http://www.rapidonline.com/netalogue/specs/81-0142.pdf

Best regards,

Luciano

[The Master]

Thanx for pointing that one out. Its nice and cheap too.

I do still have a few questions though. I dont want to just use the one you picked, i want to understand why you picked that one.

I see in the datasheet Collector-Emitter Voltage: 30V and Collector Current: 500mA so i guess thats one of the first things i should check. Both of those value must be equal to or greater than the requirement for the load.

Is it the Base-Emitter On Voltage that says how many volts you need to put into the base to make it turn fully on?

Why did you use the 10K and 100K resistors? The 100K looks to be a simple grounding resistor that keeps the base grounded while there is no input from the PIC. The PIC im using (and i thought all pics) grounded output pins when you set them to low. In this case would i still need that resistor?

Where you have put the 10K i normally have a 1K or nothing at all. I assume its to limit the current that gets to the base but how do you know what value resistor to use?

[Luciano]

Hi,

The MPSA14 is a Darlington transistor. (See datasheet page 1).

http://www.rapidonline.com/netalogue/specs/81-0142.pdf


The Darlington transistor (often called a Darlington pair) is a semiconductor
device which combines two bipolar transistors in a single device so that the
current amplified by the first is amplified further by the second. This
configuration gives a high current gain and can take less space than two
separate transistors because the two transistors can use a shared collector.

* * *

The 100k resistor can be omitted if you can afford to have the LEDs in an
unpredictable state during the initialization of the microcontroller.

The 10k resistor must be present.

Best regards,

Luciano

[The Master]

That makes sense. I wont need the 100K resistor then. It takes less than a second for the PIC to do whatever it needs to do on startup and if all the LEDs flash slightly as the PIC turns on then it doesnt really matter. I was reading something in the PIC datasheets about which pins are in what state when it first turns on so i guess that will affect it but its still not a problem. The circuits are going to be designed so all the LEDs can be on at the same time so theres no danger of using too much current.

Thanx for everyones help!

[rmteo]

If you need more than 1 transistor, for about 0.50c you can get 7 darlingtons (50V/500mA, diodes etc.) in a 16-pin package:
http://focus.ti.com/docs/prod/folders/print/uln2003ai.html

[skimask]

If you need more than 1 transistor, for about 0.50c you can get 7 darlingtons (50V/500mA, diodes etc.) in a 16-pin package:
http://focus.ti.com/docs/prod/folders/print/uln2003ai.html

But if you look at Figure 15 of the datasheet, you'll see you're limited to a little over 100mA of collector current across all 7 outputs at 100% duty cycle.
And besides, he doesn't have 'the room' for a 16PDIP or TO-220's.
Sounds like it's time for a redesign...

[The Master]

Usually when i have the need for transistors i also have the need to dot them around the PCB so they can be squeezed into small areas. Having an array of them in a DIL package may be smaller in all but it cant be split up to fit into the little spaces i have spare. As it happens ive redesigned this circuit and all the transistors are in a line but i do only need 4 of them.

Normally a TO220 package wouldnt be a problem especially if i need a lot of current but this time i only need 200mA and i dont have the vertical space for them. There will be 2 PCBs that are only 10mm apart. One of the voltage regulators will be soldered to the underside of the PCB and both of them will be folded over to be parallel with the PCB.

There will be a bigger version of this light with more LEDs but for this one im limited with space. Ive redesigned the circuit to use 4 of those transistors Luciano recommended and they all fit on fine

[The Master]

I just noticed something. Do you mean MPSA14 or MPSA13? The datasheet you posted is for MPSA13 and thats what ive been reading. The order code you posted is for MPSA14 though. In the datasheet for the MPSA14 it says Ic is 1.2A. The datasheet ive been looking at says that Ic means collector current and that one is only 500mA. Both look suitable to me except the MPSA13 is a little cheaper.

[Luciano]

The datasheet you posted is for MPSA13 and thats what ive been reading.
The order code you posted is for MPSA14 though.

Hi,

The datasheet I have posted is the datasheet of the MPSA13 and the MPSA14.
http://www.rapidonline.com/netalogue/specs/81-0142.pdf

The first page of the datasheet is added by Rapid. If you scroll to page 2
you will see the real page 1 of the datasheet and on the top right corner
you can read MPSA13 and MPSA14.

The collector current for the MPSA13 and the MPSA14 is 500mA.
The difference is the current gain.

Time to build a prototype, try both transistors.

Best regards,

Luciano

[The Master]

Im not sure which link im looking at then. The one i got using the order code from your schematic looks completely different. It has chinese in it. http://www.rapidonline.com/netalogue/specs/47-5512.pdf. Thats the one i got the 1.2A from. I guess getting both cant hurt because the bigger versions of these lights will have more LEDs so i guess its worth testing both

[Luciano]

Hi,

Will be the transitor better than the baby milk?

Best regards,

Luciano

[The Master]

Whats that mean?

[Luciano]

Bitter taste over China baby milk
http://news.bbc.co.uk/2/hi/asia-pacific/7620812.stm

[The Master]

Ok, so china has contaminated milk powder. Are you saying not to trust the transistor with chinese in the datasheet incase its substandard or did i completely miss the point?

"Will be the transitor better than the baby milk?"
Erm, i guess it will be because i dont think baby milk will turn LEDs on

[rmteo]

Get the manufacturer's datasheet here:
http://www.fairchildsemi.com/ds/MP/MPSA13.pdf

[The Master]

So the one that Luciano origionally linked to is for both of the 2 and says 500mA for the Collector Current. Ive just had a look at that datasheet and the one for MPSA14 from the same site and both of those say the Collector Current is 1.2A.

Have these transistors been upgraded at some point and rapid still has an old datasheet?

[skimask]

It has chinese in it.
Are you sure that's chinese? Or is it a manufacturer's logo? :D

[The Master]

Im not sure exactly. It could be japanese kanji. The datasheet Luciano linked to didnt have any of that. It says its for both but the second one has a completely different datasheet. Now the datasheets from www.fairchildsemi.com are completely different to both of them :O

[Luciano]

Hi,

If the quality of the datasheet reflects the quality of the product....
http://www.rapidonline.com/netalogue/specs/47-5512.pdf

Compare the above datasheet with the datasheet of the MPSA14 made by Philips Semiconductors.
http://www.nxp.com/acrobat_download/datasheets/MPSA14_7.pdf

* * *

Make sure you know who are the makers of the components you use.

Best regards,

Luciano

[The Master]

That makes sense.

So can you get the same component (with the same number) from more than one company then? I thought each company had their own numbering system even if they make components that do the exact same thing

[rmteo]

It varies from manufacturer to manufacturer. The Fairchild one is rated at 1.2A. This one from On Semiconductor is rated at 500mA:
http://www.onsemi.com/pub_link/Collateral/MPSA13-D.PDF

[Luciano]

Hi,

"Tru Semi" is Rapid Electronics which is not a semiconductor maker.
Try the URL www.trusemi.com and you will be linked to Rapid Electronics.

Is the quality OK?
Can Rapid Electronics test the quality of these products?
Does Rapid Electronics know the real makers of these semiconductors?

* * *

Is that a commercial project or just for your personal use?

Best regards,

Luciano

[The Master]

I found something similar when i went to the truopto site. All the datasheets linked to the rapid site.

I was hoping to sell these lights if they work well. The ones im making at the moment are just test versions and will only be for personal use as will the next few versions untill im confident they are working properly

I know rapid cant test the quality of all their products but they do tend to sell good quality components. If anyone complained about anything im sure they would investigate it straight away. I suppose the only real way is to buy a few and test them. Before selling any of these lights i will leave some turned on for a week or 2 constantly. Some doing PWM and some with all 4 colors on full brigthness. Might even leave some running in extreme conditions like hot/cold/wet etc

[Luciano]

Hi,

The only problem I can see is that your LED modules will
have to produce the same light intensity. This is going to be
important when the LED modules are mounted close to each
other and therefore you can compare the intensity of the light.
The quality of the used LEDs and components will make the difference.
Life duration of high efficiency LEDs is an important consideration.

Best regards,

Luciano

[The Master]

I think ill probably keep one of them turned off then and compare it with the ones ive been testing. Rapid say the LEDs im thinking about are very good quality, last a long time and are bright so i guess all i can do is test them to be sure.