Help with design - NPN High Power BJT

[UR_Correct]

Hey folks,

First time poster. I'll try to be as specific as possible.

We have been working on our senior design project for one semester and we have a semester remaining. We came to a snag after being very successful with our design. Essentially, we're using an arduino deumilanove micro controller. We have it outputting a pulse-width modulation (PWM) signal at 5V with proportional control (will elaborate more if needed). It does a PWM at about 500hz.

The problem is that we need to amplify this signal to power some TEC's, which take like 3A each (we might need about 4 for our purposes). We originally bought an H-bridge (for anyone unfamiliar, it essentially takes a small signal and amplifies it from some other source), and it worked alright at low currents, but whenever we got up to about 1A, it would fry. Its datasheet says it should support at least 3A. I was talking to an experienced EE and he laughed when I told him we were using an h-bridge, saying, "good luck!"

I came up with the idea to use relays. You know, the ice cube relays (not solid state) since they have complete isolation from the leads and the coil that switches the leads. This seemed like a good fix since the leads can take something like 10A and it doesn't take much current to power the coil. By taking the signal from the arduino to power the coil of the relay, I could hook the relay leads to a different source and connect one output of the relay to the TEC and leave the other output open. (Does this make sense?) This would end up doing the job that the h-bridge should have been doing.

It turns out that the arduino can only output 40mA, and the relays I used need at least 80mA to power the coil. I then thought that I could use one of those super common transistors (p2n2222a or whatever) as a switch to then be able to make the relay work on a really low signal from the microcontroller. I got that all to work with an example circuit (replacing the uC with a waveform generator so I can change amplitude and frequency) with much excitement (my partners love me right now).

FINALLY, the last snag. Like I said, the arduino is outputting this signal at around 500hz. These ice cube relays can't keep up with that quick of switching. THey work fine to around 50Hz and then the unforgiving laws of physics take over and the thing just stops switching. I then figured, well... let's scrap the relay idea and just use a really high current transistor.

My question is this: is this possible? I ordered two transistors from digikey a couple days ago:

2N6284: http://www.st.com/stonline/books/pdf/docs/5236.pdf
and
TIP142T: http://www.fairchildsemi.com/ds/TI/TIP142T.pdf

to test out this idea. Unfortunately, it didn't work. I think I'm overlooking something. My partner thinks that it has to do with the high voltage ratings on the transistor. Okay I'll outline what I want to do in the following and someone tell me if this is possible, and if so, is it possible with either of those transistors I listed above.

I want a low signal going into the base (5V, <40mA, square wave) to use as a switch, so that every time the square wave is on /at 5V, the high collector current flows through the emitter, and when the square wave is off or at 0V, the circuit is open, with no current flowing through the collector. I need a really high current through the collector, like 10A. The problem I think is that if I connect 4 TEC's in parallel (they are about 1 ohm each, so the eq R is about 0.25 ohms), that means I would need like 3VDC at the the collector side to get 12A and 3A through each. Is this possible?

So my set up right now is I have 12VDC to an 8ohm resistor (20W resistor) that goes to the collector of the TIP142T NPN BJT. This would mean that is about 1.5 A going through the collector. The datasheet for this transistor says that the current gain (hfe) is 1000 @ 5A, so I used that. That would mean I_B = 1.5A/1000 = 1.5mA. The datasheet also says that Vbe to turn on is 3V (but at a higher current) So then I have a 5V square wave @ 1kHz from a function generator connected to a R_b ([5V-3V]/1.5mA = 1.33kOhm) connected to the base pin. I connected an oscilloscope across the 8 ohm resistor, but once I turned on the 12V supply and the waveform generator, the resistor voltage reads 12V, which means it's not working (probably fried the transistor). I have another one and the 2n6284, but I don't want to try the same thing and fry those too.

So, am I doing it wrong? I'm still an undergrad, and I know there's so much I still don't know, but I was hoping that the brilliance of this forum could help me out or lead me in the direction or let me know that my design will be fruitless. Any help would be greatly appreciated.

Thanks, I know this was a long one.

PS I had a really bad electronics professor. I had to learn how to use a transistor on my own (most kids at my level still don't know either), and I still really only know how to use it as a switch. Like I said, I got the p2n2222a to work as a switch, so I know the calculations it takes to get the right resistors, currents, etc, to make it work, but my nomenclature might be off since I never learned it formally. Maybe it's just that this darlington transistor has much more obscure properties that I don't understand.

PSS, I just thought, what if I put in a delay in the code for the PWM to make it around 50 Hz? Would that be an easy fix, and I could keep using the relays?

EDIT: Schematic Attached in post #10 (sorry, I'm not good with all the forum stuff yet), and thanks to everyone for their help. For science!

 

[Zryn]

I'm not sure what a TEC is, but it appears that you have the perfect setup for a power converter (aka switch mode power supply or DC/DC converter) of some kind.

You already have a PWM control signal that could get a Gate Driver to turn a MOSFET on/off (like an electrical switch version of a relay that can operate into the 100kHz region easily) to deliver a current from a source to a load.

It might be too complicated of a solution, but if you take a plug pack ($20 ish) that turns the wall socket AC into a DC source, put that through a buck or boost converter, you could get the current/voltage you need for your load (TEC) no worries.

You just need to know info on size/time/weight/money concerns for your project.

 

[UR_Correct]

Thanks, man.

I think I"ll do a little more research on MOSFETs. Maybe BJT's aren't my best bet for this application. Do you think it's a better idea for isolated switching? I need a small signal to do the switching, and I need a very large current that is being turned on and off.

 

[berkeman]

Hey folks,

First time poster. I'll try to be as specific as possible.

We have been working on our senior design project for one semester and we have a semester remaining. We came to a snag after being very successful with our design. Essentially, we're using an arduino deumilanove micro controller. We have it outputting a pulse-width modulation (PWM) signal at 5V with proportional control (will elaborate more if needed). It does a PWM at about 500hz.

The problem is that we need to amplify this signal to power some TEC's, which take like 3A each (we might need about 4 for our purposes). We originally bought an H-bridge (for anyone unfamiliar, it essentially takes a small signal and amplifies it from some other source), and it worked alright at low currents, but whenever we got up to about 1A, it would fry. Its datasheet says it should support at least 3A. I was talking to an experienced EE and he laughed when I told him we were using an h-bridge, saying, "good luck!"

I came up with the idea to use relays. You know, the ice cube relays (not solid state) since they have complete isolation from the leads and the coil that switches the leads. This seemed like a good fix since the leads can take something like 10A and it doesn't take much current to power the coil. By taking the signal from the arduino to power the coil of the relay, I could hook the relay leads to a different source and connect one output of the relay to the TEC and leave the other output open. (Does this make sense?) This would end up doing the job that the h-bridge should have been doing.

It turns out that the arduino can only output 40mA, and the relays I used need at least 80mA to power the coil. I then thought that I could use one of those super common transistors (p2n2222a or whatever) as a switch to then be able to make the relay work on a really low signal from the microcontroller. I got that all to work with an example circuit (replacing the uC with a waveform generator so I can change amplitude and frequency) with much excitement (my partners love me right now).

FINALLY, the last snag. Like I said, the arduino is outputting this signal at around 500hz. These ice cube relays can't keep up with that quick of switching. THey work fine to around 50Hz and then the unforgiving laws of physics take over and the thing just stops switching. I then figured, well... let's scrap the relay idea and just use a really high current transistor.

My question is this: is this possible? I ordered two transistors from digikey a couple days ago:

2N6284: http://www.st.com/stonline/books/pdf/docs/5236.pdf
and
TIP142T: http://www.fairchildsemi.com/ds/TI/TIP142T.pdf

to test out this idea. Unfortunately, it didn't work. I think I'm overlooking something. My partner thinks that it has to do with the high voltage ratings on the transistor. Okay I'll outline what I want to do in the following and someone tell me if this is possible, and if so, is it possible with either of those transistors I listed above.

I want a low signal going into the base (5V, <40mA, square wave) to use as a switch, so that every time the square wave is on /at 5V, the high collector current flows through the emitter, and when the square wave is off or at 0V, the circuit is open, with no current flowing through the collector. I need a really high current through the collector, like 10A. The problem I think is that if I connect 4 TEC's in parallel (they are about 1 ohm each, so the eq R is about 0.25 ohms), that means I would need like 3VDC at the the collector side to get 12A and 3A through each. Is this possible?

So my set up right now is I have 12VDC to an 8ohm resistor (20W resistor) that goes to the collector of the TIP142T NPN BJT. This would mean that is about 1.5 A going through the collector. The datasheet for this transistor says that the current gain (hfe) is 1000 @ 5A, so I used that. That would mean I_B = 1.5A/1000 = 1.5mA. The datasheet also says that Vbe to turn on is 3V (but at a higher current) So then I have a 5V square wave @ 1kHz from a function generator connected to a R_b ([5V-3V]/1.5mA = 1.33kOhm) connected to the base pin. I connected an oscilloscope across the 8 ohm resistor, but once I turned on the 12V supply and the waveform generator, the resistor voltage reads 12V, which means it's not working (probably fried the transistor). I have another one and the 2n6284, but I don't want to try the same thing and fry those too.

So, am I doing it wrong? I'm still an undergrad, and I know there's so much I still don't know, but I was hoping that the brilliance of this forum could help me out or lead me in the direction or let me know that my design will be fruitless. Any help would be greatly appreciated.

Thanks, I know this was a long one.

PS I had a really bad electronics professor. I had to learn how to use a transistor on my own (most kids at my level still don't know either), and I still really only know how to use it as a switch. Like I said, I got the p2n2222a to work as a switch, so I know the calculations it takes to get the right resistors, currents, etc, to make it work, but my nomenclature might be off since I never learned it formally. Maybe it's just that this darlington transistor has much more obscure properties that I don't understand.

PSS, I just thought, what if I put in a delay in the code for the PWM to make it around 50 Hz? Would that be an easy fix, and I could keep using the relays?

What's a TEC? Can you post a diagram of what you are trying to do?

Welcome to the PF, BTW.

 

[Zryn]

It seems as though your uC only being able to output 5V 40mA (200mW) is not going to be able to power something that requires 3A (x4 = 12A) unless it can do so at a voltage of 200mW/12A = 17mV (through a lossless transformer), which is unlikely.

Do you have to make something yourself? Is a power supply inside the scope of the project or an accessory which doesn't affect the marking were you to purchase an off the shelf item?

I don't know what a TEC is or what it does or how you would use it, could you elaborate more? AC/DC power, Volts + Amps + Watts, variable power requirements range?

 

[Studiot]

TEC = Thermoelectric cooler?

3 amps?
You should be able to drive these at 500Hz with proper design of BJT or mosfet circuitry.

The TIP & 2N are darlingtons, which will not work without proper biasing.

As others have said, circuit design is easier with a diagram than (edit pictures)text . Can you come up with at least a block diagram?

Incidentally I would not try to interface directly, but insert an intermediate buffer to protect the PWM source in case of catastrpohic failure at the output. You are lucky you have not done that yet and only destroyed output transistors.

However crispy output transistors is part of the staple diet of budding circuit designers.

 

[UR_Correct]

My bad. A TEC is a thermoelectric cooler. It's a pretty dumb device. I'm not sure on the complete physics of it, but it consists of two plates of different material. My best analogy is that one plate is like 3x hot and the other plate is 1x hot. The amount of current you run through it (depending on the unique plate specs) induces a temperature difference or energy transfer called a Q value (measured in Watts) through the plates and is typically 60% of the power you give it. If you attach a heat sink to one side and depending on the direction of current, one side will heat or cool. The one we have is about 1 ohm. So we've been running 5VDC across it with a variable resistor in series so we can vary the current running through it (just to test out some properties for another part of the design). This one's max is about 4A. So I would estimate on average it is using like 15W or something. So 4 of them would obviously need about 60W.

Yes, the microcontroller can only output a small signal. That's why I thought relays would be a good bet. I can use the signal from the uC to power the coil of the relay so that every time the uC outputs a high signal (5V) the coil will actuate the switch. Since the leads of the relay I was using can take a pretty high current, I'll just hook an external supply (we're going to be plugging into a wall, and I already have a scheme to power it [rectifier bridge, voltage division, transformers, etc. Might just buy some already built AC/DC and voltage converter], so that's not a problem) to the input lead of the relay and connect the TEC's to the normally opened output, and keep the normally closed output of the relay just open (nothing connected to it). Keep in mind I only had DPDT (double pull double throw) relays at my disposal at the time. I know I can use SPST to do the same thing. The problem is that a relay like that can't switch fast enough (uC outputs pulses at ~500Hz). I was wondering if there is a type of transistor that can do this. I know they are used for high speed switching.

I will draw up a schematic tomorrow and maybe it will shed some light on some foggy areas.

 

[UR_Correct]

TEC = Thermoelectric cooler?

3 amps?
You should be able to drive these at 500Hz with proper design of BJT or mosfet circuitry.

The TIP & 2N are darlingtons, which will not work without proper biasing.

As others have said, circuit design is easier with a diagram than pictures. Can you come up with at least a block diagram?

Incidentally I would not try to interface directly, but insert an intermediate buffer to protect the PWM source in case of catastrpohic failure at the output. You are lucky you have not done that yet and only destroyed output transistors.

However crispy output transistors is part of the staple diet of budding circuit designers.

Yes, I surely wasn't going to hook anything up to the uC without knowing for sure it would work without, as you said, catastrophic failure. (My partners on the other hand can get a little overeager. ...10 fried h-bridges later).

I actually only recently learned how to properly design a BJT (bad electronics professor) as a switch for the purposes of this project on my own with a standard 2n3904, but those collector currents are far too small for what I needed. (The small victory I had that day was beyond satisfying) And yes, many a transistor died that day. But yeah, I then scoured through digikey looking for a transistor that could could take +10A through the collector, have it go on from a small signal (5V, 40mA), which means high current gain (~500, right?). I bought those two that I linked in the original post with no avail. Should I avoid the darlingtons? Do you have a better suggestion?

Thanks

 

[MATLABdude]

As has been alluded to in this thread, you'll probably want to use the power BJT in saturation mode (to complete a path to ground, in the case of an NPN) or a MOSFET in linear mode (again, to complete a path to ground, in the case of an NMOS).

From your first post, it sounds as if you're trying to do this (the Common Emitter configuration) but realize that you're designing for linearity (using the nominal linear $\beta$ or $h_{fe}$) instead of designing for BJT saturation (in which $\beta_{sat}< \beta$).

Unless you need to switch current direction on your TEC, you don't need an H-Bridge, just a simple transistor switch, as Studiot says. For more discussion on microcontroller switching of power electronics, you might want to take a look at this thread (which, full disclosure, I contributed to):
https://www.physicsforums.com/showthread.php?t=402838

EDIT: ...and made an IM(notso)HO, nice schematic drawing. Also, you'll probably want to put a reversed diode in parallel with the TEC, I seem to recall (perhaps incorrectly) that those guys had quite the inductive kick.

 

[UR_Correct]

As has been alluded to in this thread, you'll probably want to use the power BJT in saturation mode (to complete a path to ground, in the case of an NPN) or a MOSFET in linear mode (again, to complete a path to ground, in the case of an NMOS).

From your first post, it sounds as if you're trying to do this (the Common Emitter configuration) but realize that you're designing for linearity (using the nominal linear $\beta$ or $h_{fe}$) instead of designing for BJT saturation (in which $\beta_{sat}< \beta$).

Unless you need to switch current direction on your TEC, you don't need an H-Bridge, just a simple transistor switch, as Studiot says. For more discussion on microcontroller switching of power electronics, you might want to take a look at this thread (which, full disclosure, I contributed to):
https://www.physicsforums.com/showthread.php?t=402838

EDIT: ...and made an IM(notso)HO, nice schematic drawing. Also, you'll probably want to put a reversed diode in parallel with the TEC, I seem to recall (perhaps incorrectly) that those guys had quite the inductive kick.

I have attached a schematic to this post.

I've heard that the reversed diode in parallel with the coil for the back emf (not completely sure though)

We do have to switch current, but I've already got a working idea for that. We have another output from the uC (+5V for current one way, 0V for the other) that signals the coil of a DPDT relay that switches the polarity on the load. It works so I think we're good there.

But yeah, in the schematic, it was basically an evolution of ideas. I had the uC (theoretically) hooked straight to the relay coil to actuate it. (I've done all this testing with a waveform generator and external sources for the purposes of saving the uC from something bad that might happen) The uC wasn't giving quite enough current to actuate the relay, so I learned how to use a transistor as a switch to essentially switch the switch. Now, I'm realizing that I should just scrap the relay and get a transistor that can close it to ground from a small signal, but can handle +10A through the collector. I'm wondering is these exist, and if so, what specs do I look for on a datasheet to get the correct one? I know how to design a transistor as a switch for the simple case (2n3904).

So, any clues?

 

[Studiot]

OK

As pointed out you are trying for common emitter mode, which is fine for voltage amplification, but poor for current amplification.

You need current amplification and hence a common collector or emitter follower (Mosfet = source follower).

Your darlingtons will perform just fine at this. Further the output impedance will match the low impedance of the TEC, which you need for power transfer.

At 500HZ there should be no problem producing 3 - 5 amp pulses with this configuration.

Can you produce a circuit diagram from this info?

Emitter followers are the basic power output configuration so if you have not met these, this is a good opportunity to learn about them.

 

[MATLABdude]

OK

As pointed out you are trying for common emitter mode, which is fine for voltage amplification, but poor for current amplification.

You need current amplification and hence a common collector or emitter follower (Mosfet = source follower).

Your darlingtons will perform just fine at this. Further the output impedance will match the low impedance of the TEC, which you need for power transfer.

At 500HZ there should be no problem producing 3 - 5 amp pulses with this configuration.

Can you produce a circuit diagram from this info?

Emitter followers are the basic power output configuration so if you have not met these, this is a good opportunity to learn about them.

Hear, hear! Theory (and in-class learning) are just starting points. If you're like me, you don't truly, truly understand until you start doing, regardless of what your profs teach in class, and how well they convey the material.

That said, to the OP you don't mention what the operating voltage of your TECs are. Unless you use a second transistor to bias, you may have to go with a common emitter / source configuration (in saturation for your Darlington, or linear for your MOSFET) in order to drive your TEC at 12 or 15V or whatever your TECs use. Using the Darlingtons in an emitter follower configuration means that you're losing two diode drops from your base, which is only 5V to begin with (assuming you're driving from your microcontroller).

The inductive kick is what causes the back EMF (you may have covered this in earlier circuit analysis or electromagnetics courses):
$$V=L\frac{di}{dt}$$

The reversed diode keeps the back EMF from getting too low, and possibly causing damage to the rest of your circuit (such as your microcontroller). When driving a relay from a microcontroller (where the coil requires more than a few mA of current) I usually use a transistor to drive, and always use a flyback diode (i.e. reversed diode). This is even more important when driving larger relays that consume hundreds of mA. I had a friend whose microcontroller kept resetting as a result of his (non-flyback diode protected) relays, and yes, personally zapped a few logic chips before I discovered what they were used for.

 

[schip666!]

According to you little schematic you only need a single ended switch, not an H-bridge. I would just put a big transistor at Q1 with a good heat-sink (an N-chan enhanced MOSFET would be better than a regular BJT as it has lower on-resistance) and use it to pull down the negative terminal of your load. The TECs probably don't have any reactive issues so you don't need to bypass them like you show with the relay. I would also have some conceptual issues with running a mechanical relay at 500Hz...

I don't know why your EE poo-pooed the H-bridge, except that it's overkill. I have some running 8amp 80volt servo motors on my milling machine.

 

[vk6kro]

OK

As pointed out you are trying for common emitter mode, which is fine for voltage amplification, but poor for current amplification.

Why? Common emitter amplifiers can have very high current gains, especially with small collector loads.



Maybe you could look at solid state relays. 500 Hz pulse-width mode switching would be a lot to ask of mechanical relays. You would have to get the duty cycle right and this would be very difficult.

If you just want to control the temperature of a Peltier device, you can have a sensor on the device and apply power until the required temperature is close and then remove it or even reverse its polarity to get heating instead of cooling, for example.

This way, you would just be turning relays on or off at a leisurely pace.

Note that one side of the Peltier device has to be thermally coupled to a heatsink.