I am needing to switch 7 high power circuits in three banks, and looking for the best way to do so. Each of the 3 banks will flash independently based on an Arduino output, but I am breaking each lighting element into it's own branch of each bank (I can't shrink or break up each element any further).
I would like to use a minimal number of components in doing this. I wasn't sure if I could bridge pairs of a ULN2803 Darlington to function as three Q1's, and have 7 separate Q2's? Or if there was a more efficient way of doing this.
Note: These will be cycling states at a pretty aggressive rate for an extended amount of time, so a relay is not practical. Additionally, the clicking of a relay would be a problem.
Perhaps a TIP42C or a 2N6107 would be the proper PNP's to use? But again, for the primary bank driver, could I use a standard ULN2803A?
Bank 1:
One set of 33 LED's (2.5 amps)
One set of 33 LED's (2.5 amps)
One set of 33 LED's (2.5 amps)
Bank 2:
One set of 33 LED's (2.5 amps)
One set of 33 LED's (2.5 amps)
One set of 33 LED's (2.5 amps)
Hi,
If all your are switching is lighting, why do you need to HIGH SIDE switch, what are the benefits over simple LOW SIDE switching in this situation?
For high-side switching I'd use a p-channel MOSFET of on-resistance less than 100 milliohms, not a PNP transistor, for the 2nd stage. Both R can be 1k or so.
For low-side switching the ULN2803 isn't up to the current levels needed.
An n-channel logic-level MOSFET of on-resistance less than 100 milliohms would work.
TomGeorge:
Hi,
If all your are switching is lighting, why do you need to HIGH SIDE switch, what are the benefits over simple LOW SIDE switching in this situation?
Tom...
I am working with a lighting system made by a vendor that is common ground, multiple input (think RGB on steroids, but in this case white/amber/red). I'd love to do simple low-side switching, but this setup just isn't designed that way
Thus, to save space, I figured I could use a 2803 as the driver for a high side switch.
magnethead494:
This is what I had in mind. I just don't know if it will work?
If you use MOSFETs instead of transistors, you don't need to worry about base current required to saturate a PNP transistor. Both resistors in your first schematic can be 10k; just change the resistor for the N channel MOSFET to be between the gate and source.
You could skip the N MOSFET if you added circuitry to level shift the 5V GPIO output to a high enough voltage that the P MOSFET would turn off. Easier to just use the N MOSFET...
Base resistor of second transistor missing if using old bipolars.
Forget the antique ULN2803, we are well into the 21st Century. Use a TPIC6A595 and you only need three control pins to drive it. Whether PNP or p-FET, you do not need to bridge outputs as this can source (actually, sink) the 250 mA needed to saturate a PNP transistor to conduct 2.5 A. If in fact, you need multiples of 2.5 A, you will need FETs.
Thank you for the input. I'm not particularly skilled at SMD work, and since this is a one-off, DIP/SIP/Through-Hole components are easier and more economical to work with.
If a P-MOSFET will be more effective, then I can do that.
I was trying to find some sort of integrated IC to do this, kinda surprised the closest I could find would be a mis-purposed H-bridge chip. I thought this would be a more common situation.
dl324:
If you use MOSFETs instead of transistors, you don't need to worry about base current required to saturate a PNP transistor. Both resistors in your first schematic can be 10k; just change the resistor for the N channel MOSFET to be between the gate and source.
This diagram is helpful. For the sake of space and simplicity, what would be the difference between using a BJT or Darlington pair instead of the N channel?
I'm trying to fit this in a 4x6 potting box if I can, but I'm starting to think I'll need a bigger enclosure. Was hoping to just use an Uno and Protoshield. A pro-mini would be nice, but by the time I try to regulate the input from the 12V 20 amp SWPS, I might as well use the VIN port of the Uno. Or would the raw input of the Pro Mini have just enough capacity to do what I am asking without overheating the onboard regulator?
Other option would be to take apart a $2 cig lighter to USB adapter and use that to drive a pro mini
magnethead494:
Thank you for the input. I'm not particularly skilled at SMD work, and since this is a one-off, DIP/SIP/Through-Hole components are easier and more economical to work with.
Then no complaints about how much space it takes up.
magnethead494:
I was trying to find some sort of integrated IC to do this, kinda surprised the closest I could find would be a mis-purposed H-bridge chip. I thought this would be a more common situation.
If not for the 2.5 A requirement, an old UDN series chip would do, but - that is the point, the current.
magnethead494:
This diagram is helpful. For the sake of space and simplicity, what would be the difference between using a BJT or Darlington pair instead of the N channel?
Not much, you would require a Darlington, but - I gave you the most sensible approach!
magnethead494:
A pro-mini would be nice, but by the time I try to regulate the input from the 12V 20 amp SWPS, I might as well use the VIN port of the UNO. Or would the raw input of the Pro Mini have just enough capacity to do what I am asking without overheating the onboard regulator?
What makes you think the regulator on the Pro Mini is different to that on the UNO?
magnethead494:
Other option would be to take apart a $2 cig lighter to USB adapter and use that to drive a pro mini
Common recommendation here.
But there are plenty of tiny and well-built buck converters available.
Paul__B:
Not much, you would require a Darlington, but - I gave you the most sensible approach!
Not saying you're wrong.
3 things at hand - 1) I'm not spending my own money 2) This is for a retail display 3) I suck at SMD work
My Realistic options, as I understand, are to either use 7 N-channel Mosfets and 7 P-channel Mosfets (14 components, 42 pins), or I can use a 2803 and 7 P-channel mosfets (8 components, 39 pins).
In my mind, the 2803 route would be more condensed and less redundant.
magnethead494:
This diagram is helpful. For the sake of space and simplicity, what would be the difference between using a BJT or Darlington pair instead of the N channel?
You could use a general purpose NPN transistor, such as 2N3904, instead of the 2N7000. With R2=10k, you can use 10k in series with the base; that will provide about twice as much base current as you need.
The rule of thumb is that you assume a beta of 10 when using the transistor in saturation mode (some transistors such as BC547 specify a beta of 20). With a 12V supply:
Ic = 12V/10k ohms = 1.2mA
Rb = (3.3V - 0.7V)/0.12mA = 21.7k
ULN2803 were designed to interface directly with 5V logic signals so you could use them without base resistors. They're overkill for your application, but will save you board space.
What would an ideal Q2 be? Something like an FQP7P06 ? This particular datasheet does not indicate that the tab is connected to any of the pins (I know that's a thing to be careful of on BJT's).
Since I'll be moving 2.5 amps (30 watts) through it, I assume a heatsink of some sort would be a good idea? If the tab is in fact electrically isolated, I could just run an aluminum bar across the tabs.
magnethead494:
What would an ideal Q2 be? Something like an FQP7P06 ?
It'll do the job. You could look for one with a lower on resistance and Id. It would be uncommon for the source or drain to not be connected to the tab.
Since I'll be moving 2.5 amps (30 watts) through it, I assume a heatsink of some sort would be a good idea? If the tab is in fact electrically isolated, I could just run an aluminum bar across the tabs.
You're using the MOSFET as a switch. A typical FQP7P06 would have an Rds(on) of about 350m ohms at 2.5A. Power dissipation at 2.5A would be (2.5A)(2.5A)(350m ohms) = 2.2W. A heatsink is still needed.
dl324:
It'll do the job. You could look for one with a lower on resistance and Id. It would be uncommon for the source or drain to not be connected to the tab.You're using the MOSFET as a switch. A typical FQP7P06 would have an Rds(on) of about 350m ohms at 2.5A. Power dissipation at 2.5A would be (2.5A)(2.5A)(350m ohms) = 2.2W. A heatsink is still needed.
Thanks. So i'll need to heatsink them individually somehow without allowing them to touch eachother, I take it. (I still find it odd the datasheet doesn't specify). Or use nylon nut and bolt and thermal pads.
magnethead494:
I still find it odd the datasheet doesn't specify
You are right! The datasheet is missing that critical detail. The tab is almost always the same as the middle pin on any TO-220 package; they are presumably punched in one piece.
Mind you, that datasheet is actually from 2001 (down near the bottom). That really is a rather poor device for your application, you would be much better off using a modern part requiring no heatsink. If it is a SMD part, just use these adapters:
The chip plus the adapter will be no bigger than the TO-220 itself. You will have a vastly more compact assembly.
Paul__B:
You are right! The datasheet is missing that critical detail. The tab is almost always the same as the middle pin on any TO-220 package; they are presumably punched in one piece.
Mind you, that datasheet is actually from 2001 (down near the bottom). That really is a rather poor device for your application, you would be much better off using a modern part requiring no heatsink. If it is a SMD part, just use these adapters:
The chip plus the adapter will be no bigger than the TO-220 itself. You will have a vastly more compact assembly.
What would be a better P-channel Mosfet? I've only done a few pieces in SMD and none of them were exactly pretty; granted 3 pins on an 0.1 spacing is easier than 10 pins on a .0254 spacing.