After realizing the l293d output won't suffice and the voltage drop of the l298 is quite high, I believe learning to construct a simple H-Bridge to suit would be beneficial. Before starting with tests, am I correct in assuming the voltage drop of the below circuit would be calculated by the data sheets Rds(On) of each mosfet in the high/low path multiplied by drain source current? If so, a theoretical n/p-channel pair with Rdson .4ohms with 2A would be a .8V drop. Round to 1V. Are there other obvious factors I am ignorant of that would affect voltage drop available to motor besides thermal characteristics? Thanksenter image description here

  • 1
    \$\begingroup\$ Impedance ratios are critical for fast slew rates and efficiency. Consider motor Rg/RdsOn/DCR (Motor) make each high current ratio roughly 100. otherwise add more stages. the high current needs to drive the load Ciss on the Gate at fast slew rates and the output stage needs to drive DCR starting the motor, with 10x rated load current. So what I am saying is use a switch impedance ratio of 100 ( to 200) \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 May 22 '17 at 5:50
  • \$\begingroup\$ Thanks for pointing out tips regarding impedance ratios. I will move study into that area for a better understanding. \$\endgroup\$ – Archaeus May 22 '17 at 13:07

You are correct about the voltage drops in the mosfets. The only other major factor would be your supply (Vdd) and its ability to source the current without dropping (e.g. if the wire between the bridge and supply is long and or small). The other thing to pay attention to when using the Rds(On) from the datasheet is the gate voltage (Vgs) that Rds(On) is specified at. If the Vgs is sufficiently different (i.e. the Vgs = 10V and you're going to drive it with 5V) then you should look at the curves to see roughly what Rds you'll get for that gate voltage.

Also, most MOSFETS have built in diodes (called a body diode) so adding the extra schottky diodes isn't usually necessary.

  • 3
    \$\begingroup\$ The body diode of MOSFETs vary from slow and takes forever to "rip out" the majority carriers from the substrate to state of the art. However, if the Vdd is low enough to allow Schottky in the first place, I would always go for parallel Shottkys to have lower Vf and thus avoiding pushing current though the body diodes in the first place and thus potentially avoid any problems that might lead to in the future. If you don't need them, leave them unmounted later on. \$\endgroup\$ – winny May 22 '17 at 9:14
  • \$\begingroup\$ I will keep that in mind. Cheers. \$\endgroup\$ – Archaeus May 22 '17 at 13:21
  • \$\begingroup\$ I've stopped using extra diodes for sometime now without issue. However, I always use a large fast TVS in parallel with my bridges to clamp the rail down to a safe level for the mosfets max Vds. This is important if your motor will ever do negative work. Negative work happens when moving in the opposite direction of the applied torque and will happen whenever you try and reverse the motor direction. \$\endgroup\$ – kkemper May 22 '17 at 21:01

Don't use resistors!!! There are gate drivers for that. Driving a gate takes sometimea 5A, definitely must be done quickly, not through RC filter (C is the gate). Otherwise you can't do PWM and if you just switch, you risk burning a MOSFET by heat.

  • \$\begingroup\$ The danger is entirely dependant on the load, Vdd, and relative mosfet size. If Vdd is small then you can take your sweet time \$\endgroup\$ – kkemper May 22 '17 at 5:08
  • \$\begingroup\$ Sure. But generally this is a bad idea to use resistor. Maybe a small one if you actually don't use fst switching. But in Hbridge... \$\endgroup\$ – Gregory Kornblum May 22 '17 at 5:14
  • \$\begingroup\$ (sorry my previous comment was not complete) The danger is entirely dependant on the load, Vdd, and relative mosfet size. If Vdd is small then you can take your sweet time turning it on. Likewise if the load is small (high impedance) then you can take plenty of time. It's usually when the load is very low resistance and the rail very high that the turn on time becomes an issue. If they're working with an L293 then I don't think we're talking about a high performance motor controller. \$\endgroup\$ – kkemper May 22 '17 at 5:15
  • \$\begingroup\$ Obviously, if they would make a high-performance motor drive, they would have a specialist for that kind of questions. Yet, if his current is 2A and his voltage is 24V, power on MOSFET may reach 25W for the time it takes to open it. Well... Even a single event like this, if lasts for several milliseconds would kill it. Not to mention PWM again. It may be a low-cost low-performance motor drive, but it will still run on certain frequency. \$\endgroup\$ – Gregory Kornblum May 22 '17 at 9:15
  • 2
    \$\begingroup\$ What is the load? What is the frequency? Most importantly, touch the MOSFET, be sure it's not hot. I had a client who's MOSFETs overheated with a 10R resistors. They didn't believe it can be the reason until they shorted them. \$\endgroup\$ – Gregory Kornblum May 22 '17 at 13:31

As you have the schematic now, you will have high switching losses. Things to consider are:

  • Gate driving. You do not want the mosfet to stay in linear region with high Rdson for a longer time than needed. Especially if you want to apply PWM on the gate.
    You want to PWM the current, not the gate voltage. Use gate drivers, eg: MC33883.
    The resistors around the gates in your current diagram will slow down the switch time.

  • back-EMF of the motor. When you turn off motor current, there will be current coming back out of the motor due the inductance and magnetic properties of the motor. When this current has nowhere to go, it will create a high voltage capable of destroying the mosfets.
    To handle this you could rely on the internal body diode of the mosfet. But if you're dealing with a large motor, additional diodes might be required.
    Note: mosfets fail shorted.

  • Heat, the mosfets will become warm or hot. You're looking at 2A with 1V drop, that's 2 Watts. Don't forget the body diode current in the heat calculation. Make sure you do not fry the mosfets.

A benefit with using gate drivers is that they often have charge pumps allowing you to use only N-channels. Which have better Rdson == less heat.

  • \$\begingroup\$ I have tested the motor with hex Schmitt pwm and one MOSFET. I had overheating issues at first, but solved it. It runs well and cool. Adding a p-channel to the mix with a slightly higher Rdson might introduce new issues. I might look into picking up a gate driver to eliminate road blocks. Thanks. \$\endgroup\$ – Archaeus May 22 '17 at 13:20

Do not install the capacitor EMI, unless you install one additional choke at the output. With the capacitor installed, a lot of AC (PWM switching ) current will pass through it. The capacitor, you mention is for suppression of EMI due to brush commutated rotor current that is fed with DC voltage, already filtered and smooth.

However you can use the EMI capacitor, if you put a series inductor/choke which will block the AC current. A choke that at given PWM frequency has the reactance producing cca. 1%-2% voltage drop is enough.

  • \$\begingroup\$ Glad you pointed that out with solid explanation. I wasn't planning on using the cap in this schematic anyways. Cheers \$\endgroup\$ – Archaeus May 22 '17 at 13:11
  • \$\begingroup\$ The above is extremely important and include the impedance ratios of Z(f) and RdsOn to understand where/how the best low Q = resonance occurs when switching with series RL and shunt C in addition to loss ratios of load/source. Then you can achieve closer to "impedance matched filter" performance, otherwise, ringing losses. An RLC nomograph speeds up this understanding for Q and L, and Coss, shunt Crf filter. Ferrite beads also help. Twisted pairs also have an impedance and help reduce EMI. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 May 22 '17 at 13:28

Adding a little bit to the other answers...

If your motor draws 2A nominal, then it will draw a lot more when stuck/stalled/starting. It would not be unreasonable to expect 10A in a stall. Brushed DC motors have tons or torque but the price is high current current.

Now, this is trouble, as high-RdsON FETs will reduce your torque (by reducing the available current) but also dissipate lots of power on motor starts or stalls.

Since your PWM will be slow by modern switching converter standards (like 25 kHz, not 250k) you do not need state of the art switching speed, therefore you can use bigger (ie, slower, but less RdsON) FETs than in a much faster switching DCDC.

Since your voltage is low, you will easily find FETs with a RdsON below 10 mOhm, which will solve your problem.

Note that your NMOS/PMOS scheme requires proper drivers... and bulletproof dead time! Don't switch both FETs at the same time!

If you want to go all-NMOS, there are integrated bridge drivers which will drive your 4 NMOS, with onboard charge pump to generate the gate voltage, and... current limiting... always nice to have current limiting in case a screwdriver finds itself across the output terminals...

There are also beefier integrated H-Bridge chips.

L9958 DRV8829

  • \$\begingroup\$ Right, deadtime! Also by proper gate driver. \$\endgroup\$ – Gregory Kornblum May 22 '17 at 19:08
  • \$\begingroup\$ @peufeu Thank you. I have run this motor with a homemade pwm and 1 55v 29A mosfet. Runs cool even after start/stop abuse for a half hour. I will most likely invest in a MOSFET driver, but I do like a challenge (even if it smokes). Question. I see that any p-channels available in my shop with similar specs still have higher Rdson, gate charge etc.. Would it be foolish to put the pwm current on the corresponding n-channel and only switching p-channel to logic low for duration of motor activation? Would this help to alleviate out of sync switching times or am I just out to lunch? \$\endgroup\$ – Archaeus May 22 '17 at 20:03
  • \$\begingroup\$ Well in this schematic both top MOS are P channel so if you want to send current into the motor, one of them has to be on... PMOS are worse than NMOS due to physics, they will never get close, although these days NMOS are so insanely good than even an inferior PMOS can do the job too... \$\endgroup\$ – peufeu May 22 '17 at 20:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.