H-Bridge for comment

[epineh]

Hi, I have a friend who has designed a H-Bridge for use as a servo drive, and was just wondering if anybody interested (and more knowledgeable than us ) could have a quick look and see if there is anything bad in the design before we get a couple of boards done.

This is the H-Bridge part only and we are looking into micro based "brains" at a later stage, for now I want to use it with a Pluto-P card and EMC for my router, and to test the design in the real world (of hobby routers )

Any tips appreciated.

Russell.

[kreutz]

Originally Posted by epineh Hi, I have a friend who has designed a H-Bridge for use as a servo drive, and was just wondering if anybody interested (and more knowledgeable than us ) could have a quick look and see if there is anything bad in the design before we get a couple of boards done. This is the H-Bridge part only and we are looking into micro based "brains" at a later stage, for now I want to use it with a Pluto-P card and EMC for my router, and to test the design in the real world (of hobby routers ) Any tips appreciated. Russell.

Russel;

C4 and C18 are too big for boostrap capacitors (value depends on the time and rate of turn on of the lower mosfets) , also GRNDA connections need to be eliminated, since the net is the same for both half legs and that will short circuit the motor terminals, you don't need a ground connection there. The mosfets have their own internal diodes so the external 4 fast diodes could be eliminated too. Later I will check the characteristics of the diodes in the gate drive circuit and boostrap circuit.

Where is the current sensor and current limit/regulation circuit? what voltage and currents are you aiming for?. What driving mode are you going to use ? You will probably need a dv/dt limiting circuit across your motor terminals also.

Regards,

Kreutz

[Mariss Freimanis]

Russel,

Numerous problems:

1) The external drain to source diodes will do nothing. The MOSFET intrinsic diodes will be at the higher MOSFET temperature making their forward voltages less than the external diodes. They will conduct most of the current leaving nothing for the external diodes to do.

2) The 2531 optos have asymmetrical on/off times (0.3uS/4uS). This will cause shoot-thru currents in your bridge when Lin and Hin are both high.

3) An IR app-note describes problems if Vbb has an excessive Vdt. Limit the bootstrap diode with a small series resistor. Another app-note mentions problems if Vee (bootstrap gnd) isn't at supply ground potential on start-up. Use a 100K resistor to gnd on each half-bridge output.

4) Obvious problems on the 2531 outputs. Your pull-up resistors are mis connected.

5) Give thought to using a single H/L input and /SD half bridge driver like the IR2104. If for some reason (Murphy's Law) both Hin and Lin go high on your circuit, MOSFET destruction is assured. This can't happen with an H/L input and you can still shut down both MOSFETs by using the /SD input.

Second, the IR2104 has about 125mA source current and 250mA sink current into the gate of your MOSFET. This eliminates the need for a gate discharge diode. The gate automatically discharges faster than it charges and there is a 500nS shoot-thru current lockout delay as well.

6) There is no need for a gate to source zener or any other diode on your MOSFETs. Vgs always is bounded by you bootstrap Vss and Vdd. Miller Effect capacitance (Cdg) prevents the voltage going negative and also the MOSFET Vgs is rated at +/- 20V.

7) Figure on having a low inductance, low ESR bypass capacitor across your MOSFET bridge supply. Reverse recovery currents thru the intrinsic MOSFET diodes require a very low power supply impedance.

8) Be careful in your board layout around the MOSFET source to ground area. It is vital this region be of as low inductance as possible. Use a "copper pour" layout, or if that's not practical, use the fattest traces you can fit. Think "If I see FR-4 green on my pcb in the power section then my traces aren't fat enough."

Mariss

[bunalmis]

I don't share your opinion for external parallel diode.

An external fast recovery diode may be connected from drain to source. This improve to recovery time. Dynamic character of external reverse diode very importand.

Parallel connected fast external diode go on before the internal mos diode if the mounted to same heatsink with power mos.

[epineh]

Cheers guys, I will pass all of this on, and answer what I know about.

The thinking of the external drain/source diodes was that the 540's internal diodes were too slow, is this not the case?

No current limit for this version, I believe he is doing the current limit on another board later on, or possibly modifying this one, again later on.

Driving mode will be PWM from the Pluto card, or micro based card later on.

All points will be taken in and thanks for the advice, this isn't exactly either of our field of expertise so it is great to get tips from those in the "know"

The intention is to get this working for our needs and post the finished board layout's and schematics here on the zone for others if they can use it.

Cheers.

Russell.

[kreutz]

Originally Posted by bunalmis I don't share your opinion for external parallel diode. An external fast recovery diode may be connected from drain to source. This improve to recovery time. Dynamic character of external reverse diode very importand. Parallel connected fast external diode go on before the internal mos diode if the mounted to same heatsink with power mos.

There is always a temperature differential between the die and the heatsink, the internal Mosfet diode is going to be always hotter than the external one, so the current sharing is going to be biased to the internal diode. The external diode will conduct during the time interval between its own reverse recovery time and the internal mosfet diode's, so it won't heat up.

[Mariss Freimanis]

Bunalmis,

My "opinion" doesn't matter, what matters is what the circuit actually does. I have tried external diodes drain to source, current probed them and measured less than 10% of the drain current shunted thru them.

Our circuitry has the MOSFET 'on' during reverse polarity (current flowing source to drain). This helps limit power power dissipation (7A^2 * 0.04 Ohms versus 7A * 0.85V or 2W vs. 6W). The intrinsic diode doesn't conduct except for the 500nS dead-time period.

Going straight to the datasheets, for a 25C Tj, an IRF540N Vsd is 0.85V at 7A. The BYV27 shows an If of 0.7A for a Vf of 0.85V at the same junction temperature. That matches observed scope measurements; 90% of the current passes thru the MOSFET, 10% thru the external diode. Things only get worse as the MOSFET heats up; MOSFET Tj will always be higher than the diode Tj.

There are circuit techniques to shut off the MOSFET intrinsic diode but they are messy and expensive. Place a Schottky rectifier in series with the MOSFET, cathode to the drain. Connect your high-speed Si diode anode to MOSFET source, cathode to Schottky anode. All reverse current now flows thru the external Si diode but you also have an additional Vf of about 0.5V per MOSFET. What was a Vds of 0.28V becomes 0.78V (Vf + Vds). Dissipation effectively triples. You also go from 4 power devices per full bridge to 12 power devices.

Just use the MOSFET's intrinsic diode; it's not so bad. We measure a trr of 100nS with an average Irr of 12A. At an 80VDC supply the pulse energy is 96 micro Joules. At a 20kHz switching frequency the switching losses per bridge are 3.8 Watts.

Mariss

[samco]

This is why this bridge was made the way it was. (low inductance + cap right across the supply)

http://www.electronicsam.com/images/...ostart/top.JPG
http://www.electronicsam.com/images/...art/almost.JPG
http://www.electronicsam.com/images/...t/mounting.JPG
http://www.electronicsam.com/images/...rt/ampmess.JPG

I have not blown one up yet - but I have not pushed it to where I want to be - 150v 20a just 60v 10a so far.

[acondit]

Russell,

What's happening? Did you go back to the drawing board?

Alan

[epineh]

Hey Alan, well after looking at going back to the drawing board, we have decided to use Samco's design (with his permission) with a few minor changes, after all his design is tested, up and running.

Main changes are to change the gate driver to an IR 2104 which will also give us an enable input, and with that an extra opto is needed.

We will look at FET selection, at the moment the thinking is an IRFT4710.

As for the freewheeling diodes, the tracks are already there for them, so they can be put in or left out, we won't take them out of the board layout. I simply do not know enough one way or another to decide, but I do respect the advice of the knowledgable folk that have contributed...thanks again

Once the layout is done we will post the finished (so far) result.

Cheers.

Russell.

[epineh]

Ok here is the latest incarnation, hope everything is where it should be...

Cheers.

Russell.

[niclatrique]

Hi Epineh,

I work on a design similar to this one. I'm not very familiar with component selection and calculation. So, Would you mind to tell me the values of the surrounding components of your IR2104 caps and resistors.

Thanks a lot

Nic