My vehicle flasher project is progressing. Next up is configuring and optimizing the output stages.
In the attached circuit, the green area can be assumed to output a regulated 5 volts going to the Arduino Nano's 5V or possibly regulated 7 volts going to the Arduino Nano's VIN.
Everything in the red and purple areas are inviolate; these are existing circuits which cannot be modified, and hooking up to them is essentially required to be as indicated.
The area of interest at the moment is in the blue area. Each of the final three output stages should be able to drive up to 100 watts (it's unlikely to ever have to drive this level of power, but should be able to). While doing something like this is reasonably well established and not particularly new, I'm learning as I go and am interested in the whys and hows of doing things like this. So, any and all suggestions on why and how to improve performance and efficiency - including replacing with different components and/or removing/adding elements - are welcomed.
The only thing I would suggest is to lower the gate pullup resistor to 1k.
This fet has a fairly large gate capacitance, and you don't want it to switch off slowly with a 100watt load.
The fet will need a small heatsink for 100watt.
Don't forget fuses in your final build.
Leo..
R1c and R1b are rather high - certainly too high for the load to be PWM'd. The input capacitance of
MOSFET gates is high and this means you want to push or pull largish currents to switch high powers,
so the time to switch (when the device dissipates massively) is as short as possible. Try 1k or 560 ohms
(any lower and R1c will need to have a larger power dissipation than the standard 1/3W.)
I would leave R1b 10k.
Deep saturation of Q1a could make it slower to recover.
Pulldown current is already high anyway.
The only slow thing is the pullup current.
Leo..
Again, you guys are awesome; exactly the type of answers/suggestions I was looking for.
My largest concern is that these outputs will always be driven via PWM pins, and the transistors are used as pure switches (i.e. they're always both either full-on, or full-off). From what I've been able to learn, transistor switching speed and saturation are the most critical for efficiency; i.e. as MarkT suggested: maximize the speed with which both Q1a and Q1b can be fully de/saturated. I was kind of wondering about using 10K on both transistors and was hoping that lower would at least work on R1c.
I'm open to using different transistors for either; only critical part is the "high" voltage side needs to be able to switch - at most - 100 watts. If some other combination of transistors and/or supporting components can work better at PWM control, I'm all ears.
Questions:
At what power level is the "small heat sink" on the MOSFET required? 60 watts? 80? (I've yet to determine exactly how much power they're going to be driving, so good to know when I need to add heat sinks.) I'm also thinking that the faster the MOSFETs can be switched, the less need there'd be for heat sinks?
Is 1/4watt OK for R1c at 1k?
What other options exist to maximize switching speeds and saturation while minimizing the current requirement to support them?
While researching how to increase switching speed of Q1b, I ran across this. As a result of that, and of trying to reduce switching current requirements, I've altered R1c to 2.2k, and added a diode (D1a) across it (as per attached). Comments?
Thank you for your comments, Leo. I suspected that the diode across Q1b wouldn't work - I just couldn't figure out why. Didn't know enough to consider the consequences at Q1a.
In the mean time, I've been doing some more reading (I know, dangerous :o ) and found this and this, describing ways to improve switching times for Q1a. Attached is revised circuit with D1a removed, and added D1b and C1. This one makes a bit more sense - though I'm not entirely sure yet exactly why. (For some reason, I'm thinking that including D1b might allow a reduction in R1b to maybe 4.7k?)
Thanks also for the MOSFET heat dissipation estimate; seems I was close: 14.5V at 6A = 87W, so I'd likely add a heat sink around 60W just to be safe.
The calculations for most of this is beyond me at the moment. I have no idea how to determine switching losses; PWM will be whatever the Nano V3 uses. If you meant, how often will my sketch turn pin 9, 10, or 11 on or off, frequency of actual on/off will be in the neighbourhood of 2Hz. However, PWM rates will be increased during "turn on" and decreased during "turn off" (i.e. I'll be turning output stages on and off at 2Hz but turning them on and off slowly (by varying PWM levels in my sketch). Turn on/off will be set to take about .2 seconds, on- and off-time about .5 seconds.
Besides working on an actual project, a lot of my questions, changes, etc. are in the interests of learning about general practical applications, and learning about the hows and whys of various ways of implementing this kind of thing. Hence, my exploring other ways of doing things.
On reading a bit more, it appears I may have D1b backwards. Also, it may be that it will not work to prevent Q1a from saturating in this configuration.
It seems that the "speed up" capacitor C1 should do what it's supposed to, though.
Anyone with ideas on preventing Q1a from saturating (apparently one of the largest reasons for slow shut-off)? Or ideas on speeding up either or both Q1a and Q1b? Different transistors? (If suggesting a replacement for the MOSFET, keep in mind I anticipate needing to be able to switch up to 100 watts. If something significantly more suitable as a high-side switch but with lower wattage rating is available, lemme know.) It would be nice if N-channel MOSFETs weren't so onerous to implement as high-side switches...
D1b is called a "Baker clamp", but the diode is drawn the wrong way.
I think it's better to lower R1b to 1k (with a baker clamp) than adding C1.
D1b is drawn as schottky diode, but you list a normal 1N4148 diode.
It should be a schottky diode, e.g. 1N5819.
You are trying to speed up an already fast part, active fet turn-on.
And forgetting about the slow part, resistive turn-off R1c.
You could replace the transistor and all supporting parts with a 2N7000.
Leo..
Wawa:
I would leave R1b 10k.
Deep saturation of Q1a could make it slower to recover.
Pulldown current is already high anyway.
The only slow thing is the pullup current.
Leo..
The time to desaturate a small BJT is ~2 orders of magnitude less, its irrelevant here.
We are talking about switching times, not delays come to that - we don't care about
things happening 2us late, we do care about the temperature rising in the MOSFET
because the thing is switching slowly (50us)
Mark, Leo, thank you both very much for your recent comments/suggestions.
Mark, your comments about heat build-up on the MOSFET is why I'm trying to figure out how to speed things up. I thought that perhaps speeding up the BJT would help with decreasing switching time at the MOSFET - mitigating heat production in the process.
Leo, thank you for your corrections. I already knew D1b was drawn the wrong way. Not sure how I managed to list it as a 1N4148 (moot, now). As for using a 2N7000 instead of the 2N3904 "and all supporting parts", what would be needed? Just the 2N7000? It appears the 2N7000 is prone to static triggering(?) its gate, so I've added a 100k resistor (R1b) from gate to ground.
I'd like to see the only heat dissipated by the MOSFET being due to RDS(on); I'm still thinking more can be done to speed up the MOSFET; a "speed up" cap across R1c (tentatively in circuit as C1c)? If workable, wouldn't this also allow increasing R1c to something more efficient - like 10k or higher?
Thank you again for all your contributions! I'm learning a lot.
dephwyggl:
I'm still thinking more can be done to speed up the MOSFET; a "speed up" cap across R1c (tentatively in circuit as C1c)? If workable, wouldn't this also allow increasing R1c to something more efficient - like 10k or higher?
A cap across R1C would slow switching considerably, and so would a larger value resistor.
Imagine the gate being 4volt below the source (the dangerous area).
The 1k resistor would have a pullup capacity of 4mA.
While the transistor/mosfet would have a pulldown capacity of 10-100x that value.
As said before, the slow part is pullup.
The best solution could be a mosfet driver IC (micrel ?).
Maybe someone else can point you in that direction.
Take care that the mosfet stays "off" during Arduino's bootup.
Leo..
1) I've looking around for possible MOSFET drivers, and, rather than an IC, found some circuits with discrete components. (I'm aware that it would be simpler to use an IC, but I'd be learning less.) See attached for a couple of examples of something that I'm starting to understand - though not enough to know if/how it's applicable for what I'm after in this project.
2) So far, I've been trying to figure out a circuit I could use generally for quite high power (100 watts), and then apply that to this project. However, this particular project is not likely to switch more than 25 watts per output stage. So, maybe something like an IRF9540N (instead of the IRF4905) would have less heat issues; RDS(on) is the same, but input and output capacitances are significantly less.
So, two questions:
1) If staying with 100 watt capability would something like the above driver circuit help? Significantly enough to be worthwhile?
2) If going to the IRF9540N, would it be worthwhile to incorporate something like the above driver circuit or just stay with what I've come up with so far?
The PNP in the first is the wrong way round, otherwise its a push-pull driver - basically OK, has
lots of current sourcing and sinking ability.
The second one has the PNP the right way round and uses shunt feedback in the driver transistor to
control its collector voltage swing to about 15V below the main supply, which is good. However I
would add a zener across the MOSFET gate-source to protect from overvoltage just in case. It is
sensitive to the logic voltage applied to the first transistor base as it is basically amplifying a voltage,
not switching.
You can just use the zener and a series resistor for the same effect without needing the emitter resistor,
then the circuit will work the same from 3.3V and 5V logic as the first transistor is in switching
configuration again, not a linear amplifier.
MarkT:
The second one has the PNP the right way round and uses shunt feedback in the driver transistor to
control its collector voltage swing to about 15V below the main supply, which is good. However I
would add a zener across the MOSFET gate-source to protect from overvoltage just in case. It is
sensitive to the logic voltage applied to the first transistor base as it is basically amplifying a voltage,
not switching.
You can just use the zener and a series resistor for the same effect without needing the emitter resistor,
then the circuit will work the same from 3.3V and 5V logic as the first transistor is in switching
configuration again, not a linear amplifier.
Mark, thank you tons for your reply, but you've left me entirely bamboozled. :o I'm sure it all makes sense - to someone who understands transistors and their applications a bit better. Entirely beyond me, at the moment. Would you be willing to provide a circuit sketch implementing your last para? Particularly to support the 5V PWM Nano pin output?
Also, I had intended to use a 2N7000 off the Nano to drive the MOSFET's gate; is this still possible? Also, to futher improve the MOSFET's switching speed, would replacing any the push-pull BJTs with MOSFETs (maybe BS170/BC250?) be of possible benefit? If so, what might that look like?
Cheers!
Dirk
Edit: The above should have read "(maybe BS170/BS250?)"...
Not sure if you're testing me (though I wouldn't object ) but isn't the driver MOSFET mislabeled? (You have it as a 2N7002 - which is a P-channel, but you've drawn it as an N-channel - 2N7000.) It appears to be sinking current so should be an N-channel?
A couple of substitution questions:
Can a 1N5819 be used instead of the 1N5818? If not, why not?
Can any small signal NPN BJT be used instead of the BC547B (I have a crap-load of 2N3904s...)? Or did you choose the BC547B for a specific reason? (I'm still working on deciphering specs.)
Mark, thank you tons for your reply, but you've left me entirely bamboozled. :o I'm sure it all makes sense - to someone who understands transistors and their applications a bit better. Entirely beyond me, at the moment. Would you be willing to provide a circuit sketch implementing your last para? Particularly to support the 5V PWM Nano pin output?
Also, I had intended to use a 2N7000 off the Nano to drive the MOSFET's gate; is this still possible? Also, to futher improve the MOSFET's switching speed, would replacing any the push-pull BJTs with MOSFETs (maybe BS170/BC250?) be of possible benefit? If so, what might that look like?
Cheers!
Dirk
Edit: The above should have read "(maybe BS170/BS250?)"...
You said you wanted to learn! If you don't know about feedback I'd recommend finding out more,
its everywhere in electronics. I'll see if I can sketh something
Note that the 10V zener means the p-channel gate will be pulled 10V below the source
reliably, the supply voltage can vary quite a bit and it will still work. You don't need
a gate resistor when driving a single MOSFET, in fact the driver transistors will limit the
current anyway as they only have a few mA of base drive.
The NPN pre-driver can be an n-channel MOSFET if you want, its doesn't matter so
long as it can comfortably handle the max voltage at its drain or collector. The push-
pull transistors should also have suitable voltage ratings (well only the PNP one
gets to see more than 10V)
Next up is configuring and optimizing the output stages.
Not sure how I managed to list it as a 1N4148 (moot, now). As for using a 2N7000 instead of the 2N3904 "and all supporting parts", what would be needed? Just the 2N7000? It appears the 2N7000 is prone to static triggering(?) its gate, so I've added a 100k resistor (R1b) from gate to ground.
