I have put together a h-bridge circuit for use in a radio control car, and was hoping that someone could have a look at it for me. I am worried that I am likely to blow the thing up ;) The circuit is here:
formatting link
(~13k)
or, if you prefer in pdf:
formatting link
(~27k)
The PWM signal is being generated by a PIC micro.
Specifically I would like to know if the use of the AND gates will insure that only one side of the bridge comes on, and if the transistor inverter will be a suitable inverter.
Also, is the fuse ok? (Is it likely to save the FETS in the event that I do stuff up?)
it will blow up the first time you try to use it :{
there are no base resistors on any of the bipolar transistors. When Q5 turns ON, it pulls the base of PNP Q6 to 0V. But the emitter is connected to +7V, so a great deal of current will flow, leading to the destruction of the pnp, probably followed by the NPN.
the logic outputs wont like driving into a base-emitter junction very much either, and may break. A smattering of (say) 1k base resistors would cure that.
the single-ended FET drive is set up to give fast turn on and slow (via
1k only) turn off of all 4 FETs, which is exactly the opposite of what is generally required, and usually results in shoot-thru current (both FETs on for a little bit). If the shoot-thru period is long, the FETs expire rapidly. If its short, then the high peak power dissipation (7V^2/(2*Rdson)) during shoot-thru averages out to a not-so-destructively-high value, and the FET runs hotter than you expect
- often hot enough to make it fail in 5-10 minutes (depends on thermal time constant of heatsink). low switching frequency reduces the average power.
Deadtime is often used to solve this problem - a brief (adjustable) period during which both FETs are off.
why not run the 4073 from the +7V supply, and let it level-shift the
0-5V micro outputs to 0-7V. A few gates paralleled ought to be able to directly drive your FETs (I havent looked at the fet data though)....
the FETs wont like the 10nF caps much if Fswitch is high
learn to use base resistors. 4000-series CMOS is happy to run up to 12V or more, why not use a real inverter?
that being said, I have a design that uses a BC847BPN as a +12V push-pull inverter with a 70ns propagation delay, and we've made 20,000 of them....
the AND gates are a good start, but dead-time is really required to prevent shoot-thru (aka cross-conduction).
put a cap *directly* across the supply of your H-bridge, the fuse must not be between the cap and the FETs. the idea is to reduce the inductance from the cap to the FET bridge.
the fuse itself is unlikely to protect your FETs. Check the fuse I^2t figure against I^2t for the FET (Ifsm can be used to figure out I^2t). The smallest I^2t blows first, and is usually the FET. But the fuse will prevent a fire.
I never build this type of circuit without pulse-by-pulse current limiting.
I'd say the turn-off time could cause a dissipation problem due to slow switching and the cross-conduction currents. If yuo could insert some deadtime that would help (separate drive for upper and lower FETs on each leg).
Bad habit on my part I'm afraid. I always tend to omit the base resistors when drawing circuits (I guess it is a bit obvious to me that they should be there), and tend to put them in when I draw up the pcb. It is a habit that will one day cause me to burn my house down. I will kick myself for you. Twice. Hard.
Could I implement deadtime into my micro code? Or this that asking for trouble? Would it be possible to somehow monitoring the bridge to ensure that it is off before switching the other side on, maybe through the spare and gate or directly to the micro? Incidentally, what is a suitably safe dead time period?
According to the data sheet the 4073 is able to source about 1-2 mA. If I use 3 gates in parallel that is only 6 mA. If I were to reduce the pullup/down resistors to say 50 ohms, would that be better? 50 ohms would give me about 100 mA Ic current, 33 ohms would give me about 150 mA, which is still below the maximum 200 mA allowed by the data sheet.
There is also a space issue as well. I really don't want to use 3 4073 chips if I can get away with a couple of transistors.
I included them only because they came with the motor and that is the recommended configuration. I don't have them on my current forward only single fet control, so I will scrap them here as well.
Space was my primary concern. I don't have much room to play with, and it seemed a bit wasteful to include six inverters and only use one. I trust this is ok?
I don't seem to be able to find that data on jaycar's website. I (naively) assumed that since the fets could handle 40 a, that a 10a fuse would blow before the fets. If that is not the case, then I will just scarp the fuse idea altogether. Frankly, if I can't save the fets the whole thing may as well burn ;)
Thanks again for your time, I actually feel I am starting to understand (some) of it. ;)
I understand what you mean about the deadtime. What would you suggest to be a suitable period for this?
Also, are you suggesting that I drive each of the four fets individually? I had actually considered this, as it would allow me to implement a crude braking system by turning on both the bottom fets, but dismissed it because it seemed to make the possibility of shorting the fets more likely. Perhaps I should reconsider?
you can put dead-time in your software, but what happens if the software flips out? one solution is to make sure the software works. The same 55W smsp that uses the little inverter has dead-time done with schmitt triggers, series R's with diodes across them and shunt C's. Take a pulse Q, generate its complement /Q. Feed both Q and /Q into identical R//D C circuits, followed by schmitt triggers (to square the waveform up again) giving Y and /Y. Now the RC circuit sets the turn-on delay for both Y and /Y. The diodes ensure the turn-off delay is small. Voila, deadtime.
look at the time constants of charging and discharging the gates, along with Vth, to figure out the required deadtime, then add a suitable safety margin. Or look at a pair of gate voltages with a dual-trace scope, and make sure one FET is completely off (Vg = 0V) before the other one begins to turn on.
do you know how to do the RC calculations?
fair call re. 4073. Its not uncommon to use 6 inverters paralleled to drive a small FET though.
what you find with your single-ended gatedrive is that speeding it up comes at a heavy price - 50R will dissipate 7V^2/50 = 1W which is quite a bit!
try a complementary emitter follower after your (now level-shifting) 4073:
-----+------- +7V |c |/ +--| npn | |\ e [4073]--[Rbase]-+ +-----[Rgate]--- to FET | |/ e +--| pnp |\ c | -----+------- 0V
A BC547/557 combo ought to do you a few hundred milliamps, BC327/337 about an amp. use the base resistor though.
beware the voltage drop though - VGmin = Vol + Vbe. The CMOS gate will pull pretty close to 0V, so the output voltage will swing from 0.6V to
6.4V or so. If Vth is real low this can be a problem, but for your FET its about 2V, so this is fine.
SMT, use a BC847BPN dual npn + pnp in a 6-leg SOT323 (smaller than SOT23) :) Better get a microscope though.....
its fine. but without a speedup cap across the base resistor it wont be very fast.
you wont.
formatting link
Idm is 200A, which is a repetitive peak current (rep rate governed by temp rise). so its a fairly big FET. If they gave a single-shot current rating I^2t falls out directly from that number.
you also wont be able to find any I^2t data for your fuse either :)
if you were keen, you could ascertain the fuse wire size, and make a stab at I^2t based on adiabatic heating (no heat transfer) and the melting point of the metal (make a guess....)
in practice, use the smallest fuse that the equipment reliably operates with. dont use a fuse at more than 60% of its rated current, or it may fail due to thermal fatigue.
it probably doesnt matter if your circuit burns, but the building its housed in is another story.....
you may also want to ensure that fried FETs dont snot the micro - a simple series R (1-10k) can suffice there. Unless the micro is a lot cheaper than the FETs, in which case who cares.
Thanks for the detailed info Terry. I will take all your suggestions on board and come up with a revised circuit in a few days. If it is alright with everyone, I will post that up in a new thread, or if if you prefer I will stick it on the end of this one. I am not sure what the group's policy is on that?
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.