Is BJT Power H-Bridge Motor Driver Obsoleted by FET H-Bridge?

[Mechatrommer]

i feel like decades old behind with my newly built BJT (PNP/NPN) H-Bridge 5A motor driver, but since i want to save some driving transistors and design simplicity, i went with BJT way. with BJT i only need 2 driving transistors (beside the main 4X power NPN/PNP transistors in the H-Bridge), with FET H-Bridge, to get a good switching characteristic, i can think of at least:

1) 8 or 10 driving (small bjt) transistors (NFET/PFET H-Bridge config), or...
2) 4-6 driving transistors + 2 NFET driver ic (NFET/NFET config), or...
3) 4 NFET driver ic (NFET/NFET config)...

in all the FET configuration, they all seem expensive (1 driver ic ~eqv to 10 driving bjt) or otherwise need more components in the board, hence all in all more expensive... but the simpler BJT H-Bridge comes at a price of less than ideal power efficiency, ie much voltage drop ie power loss in the bridge's BJTs. did i make a valid decision? or am i just obsoleted?

[SeanB]

At 5A your BJT will not be too bad power wise. Around 200mA for the drive at a maximum, so some heat. Does the extra expense of a high side driver warrant the saving, though there you also get some nice things like charge pumps, dead time control and current sensing, along with very low drop across the active devices, and the capability to simply make it a 20A design with cheap high power devices.

[Chris C]

As long as something is useful, it's not obsolete.  Sounds like the BJT H-Bridge was useful to you, so...

I've seen a H-bridge, using 2x NFETs and 2x PFETs, and only 4x BJTs.  The BJTs are arranged as two push-pull drivers.  The output from each driver controls one NFET and one PFET.  Simultaneous conduction of the FETs is prevented by putting a resistor and diode each FET gate, in serial with the gate and parallel to each other, so that turn-off occurs faster than turn-on.  I haven't tried it, and wouldn't expect efficiency to be fantastic due to the amount of time the FETs spend in the linear region, but it's the simplest I've seen.

The BJTs can also operate as level shifters, to be driven from two lower voltage MCU pins, if arranged right.  I think the minimum part count for this, including all FETs, BJTs, diodes, and resistors, could be as low as 22 for the entire H-Bridge.

[Mechatrommer]

At 5A your BJT will not be too bad power wise. Around 200mA for the drive at a maximum, so some heat.

300-500mA to be exact. i'm using TIP41/42C as the bridge, and SS8050 as driver. not an ideal choice i guess (100V BJT to be used in 12V application ) but thats what i already have in stock. but yeah, even at the driving side... some heat going on...

Does the extra expense of a high side driver warrant the saving, though there you also get some nice things...

i know, but the early decision was made partly due to the need for smallest board footprint, as i only have bunches of TO-220 package power BJTs and FETs in stocks... serious design will need serious cost down the stream buying stocks from ebay well, maybe this early design for my quick need and preliminary learning... i hope some pregraduate kids will buy as i already have 30+ pcb ordered at hand with the wrong (swapped) pic10f206 Vdd/Vss assignment i'll need to cut and retrace each one of them

I've seen a H-bridge, using 2x NFETs and 2x PFETs, and only 4x BJTs.

i need to search this, if you have the address or keyword, that can speed things up... what i currently know is, each FET needs 2x push pull BJT.. 4x FET = 8x BJT driver. each or pair of the push pull will probably need another driver, that makes all 10x BJT, as stated above 

[C]

Way back in the SCR's time
I often saw a center tapped inductor between the two SCR with the tap to load.
When one SCR was on, just turning on the second SCR turned off the first.

Using this idea with a FET or BJT
Am I missing something here?
With the inductor resisting the current change during active region of switching, you would have less power during the switch.
The inductor would also round the edge some, lowering EMI.
By being tapped would help some with turning off other fet.

All this with a cost of some added resistance.
What would be the gains an losses with just a turn or two?

C

[T3sl4co1l]

You were not only lapped, but while restarting the race, you got stuck in a coma while industry went on into space...

The obvious answer to all this discrete BS: get an integrated chip instead!  There are ICs and modules available for modest prices (usually more than the BOM cost of a discrete solution, but not by much really, nevermind the savings on PCB area and design time), which offer control circuitry (current limiting, forward/reverse step, free run, etc.), MCU interface (I2C or SPI most often), and some serious brains all their own (piles of configuration registers to play with).

They're available over 10A and 80V and with high switching frequencies and efficiency, and that's before adding external switches.  Don't know if there are external-switch controllers, but by then you're talking HP worth anyway, and a controller / gate drivers / transistors design is hardly an imposition.

Really, a discrete solution is best left for ones' own edification.  Play with it and find what you need to find out about it, then find a chip that does everything you are now expecting.

Tim

[Chris C]

i need to search this, if you have the address or keyword, that can speed things up...

I'm certain I saw it in some semiconductor manufacturer's app note, I just can't remember which one...

But I remember the circuit.  Here's a half-bridge:



Q1, Q2, R1, R2, and D1 form a common level shifting inverter.  Input of this stage is an typical digital logic voltage, and output sources/sinks a higher voltage at a few hundred mA.

R3, R4, D2, and D3 are what cause the FETs to turn on slower than turning off; ideally producing a dead-time when switching states, when neither FET is on.  R3 is smaller than R4 because the P-FET's gate capacitance is larger.  R3 and R4 will have to be tuned if different FETs are used.  I'm not sure the values shown are optimally tuned for the FETs shown, though they did perform adequately in a quick SPICE sim.  I saw a little cross-conduction under a worst-case scenario of no load, but just a few mA.

It cannot switch voltages exceeding the FET's Vgs (18V as shown), and ideally you should keep it a few volts below.

A full-bridge is just two of these, with a total part count of 22.  It's the smallest fully-discrete solution I know.

I've left out anything not absolutely necessary.  But adding zeners to the FET gates would make it more resilient against inductive spikes caused by the load or parasitics.  And adding an anti-saturation diode (BAT85) between Q1's base and collector will eliminate a turn-off delay of a few us.

[Mechatrommer]

thanks all for input.... and thanks Chris for the circuit example thats what i need to keep my brain going...

get an integrated chip instead!  There are ICs and modules available for modest prices (usually more than the BOM cost of a discrete solution, but not by much really, nevermind the savings on PCB area and design time), which offer control circuitry (current limiting, forward/reverse step, free run, etc.), MCU interface (I2C or SPI most often), and some serious brains all their own (piles of configuration registers to play with).

as someone said... where is the fun then? what i'm trying to build is bidirectional ESC for brushed DC motor that accept PPM input from RC receiver, pretty much available in many places cheap from china. in there as well i can buy one and can make as a reference of what chip (fet driver/pwm controller etc) they are using to drive the bridge, but... several issues... even though cheap the complete ESC module, when i need more something like 5-10 ESC, the price can go up pretty quick passing the "inverse diminishing return".. i mean, now i have parts to build 10++ ESC at more affordable price, albeit less ideal and need time populating them together. if i dont need 10, i can repurpose the parts for other projects plus with building it DIY i can reprogram the mcu to my liking, currently my design not only accept PPM input like many others, but also PWM/DIR or PWM1/PWM2 input for robotics or machinery purpose, not just RC toy control, something that i havent found in the market, that was the main motivation i think, apart from learning experience. so its a dual purpose input mode ESC, thats the only thing i feel good about it

and i already have few high side NFET driver here that i want to keep for more serious project, but the cost is like 10x the small SOT23 drive transistors like SS8050. but power FET and power BJT are about the same price, so sometime its quite difficult to judge clearly, i'd rather building something that i can use right away than doing some serious economic engineering on a table. but i ask your opinion for further motivation for future direction, thanks all.

[SeanB]

12V, so you can get away using 2 P channel FETs and 2 N channel ones, 4 resistors and 2 cheap NPN transistors.  P channels have to be non logic level ones, with the resistor between 12V and gate, with the NPN pulling it to 0V to turn on, base current being limited by other resistor, or just use a logic level low power mosfet instead, you really do not care how high on resistance is, just has to pull below threshold of the pfet. Low side logic level n power fet, and if you are slow enough that body diodes are usable, which will be this case as the switching speeds will be far from stellar, you will be fine and have some EMI reduction at the expense of increase switching loss.

Can do with 4 power devices, 2 resistors and 2 jelybean logic level fets, you just need 4 output pins available, and need to make sure your firmware does the dead time control or you will have very dead mosfets in no time. For protection add 2 gate resistors and 15V zener clamps to the p channel devices, all cheap parts.

For bipolar you can do the same, just have to make the drive a little stiffer and have extra resistors in the base for current setting, and accept that the NPN devices will be darlingtons with the 2V drop, though the PNP can just be driven with 200-400mA depending on how good the PNP transistors are gain wise at the current of interest. 5A you need a gain of 50, so you might already be needing to run at 400mA just to saturate them. Mosfets are so much lower power to drive, though the bipolar will survive you going oops a bit better.

[Mechatrommer]

i believe PFET/NFET config is already way better than BJT config, where NFET/NFET is only marginal compared to PFET few subohm Rdson difference is close to negligible @ 5A imho. so PFET can live with that less few hi side NFET driver ic somemore... anyway just for the fun... my less than stellar ESC in action using substandard MMBT3904 as driver, as i'm still waiting the SS8050 part order in transit, the bridge took some heat as expected Ib 200mA, spec is >400mA, albeit smaller than 3A motor i believe...

1st picture, the BJT H-Bridge Schematics: most resistors are pull-up pull-down only. since this is BJT, i believe they are not critically needed.
2nd picture, top side
3rd picture, bottom side. populated components count for the bridge section is 14 (incl motor diodes protection)
4th picture, 2x my ESC already in action using substandard MMBT3904, they survived so far...
5th picture, resurrected damaged 4 wheeler from a nephew in law. just for the fun to the kids. i can buy them ready made RC toy. but then, where's the fun for me?

[SeanB]

Yes, entertainment for you and for the kids, and if they actually see how the toy is put together they might be more interested in the future as to how things are made and put together.