8
\$\begingroup\$

I designed a circuit (MOSFET HBRIDGE) that Should Be able to handle 70A @ 14V. Now that the board is sent out I am on to thinking about how to properly test it.

Right now the most logical way I have to test it is to take a bunch of power resistors and hook it up to a SLA 12V Battery.

I would start w/ 15A then maybe 35A then 50A or something to that extent. I would monitor the temperature on both the fets and the traces themselves.

Any other suggestions would be appreciated.

\$\endgroup\$
7
  • 4
    \$\begingroup\$ Your method sounds good to me - cheap electric kettles, heaters, or at 12v car headlamp bulbs make a good dummy load. \$\endgroup\$
    – John U
    Jun 13 '13 at 15:18
  • \$\begingroup\$ Thanks I just couldn't think of an easier way to test it. \$\endgroup\$
    – EE_PCB
    Jun 13 '13 at 15:56
  • \$\begingroup\$ I didn't say it was the best way, just a way ;) \$\endgroup\$
    – John U
    Jun 13 '13 at 16:10
  • 8
    \$\begingroup\$ I hope you're using a lot of FETs, and not just trusting the FET current ratings from the datasheets. A TO-220 package's legs will fuse at ~75A. Read this: mcmanis.com/chuck/robotics/projects/esc2/FET-power.html \$\endgroup\$ Jul 6 '13 at 19:33
  • 3
    \$\begingroup\$ Don't let the comic sans throw you... @ConnorWolf's link is an excellent discussion of MOSFET current limits. \$\endgroup\$
    – markrages
    Aug 5 '13 at 18:53
1
\$\begingroup\$

While checking temperature is good, it would be better to check for voltage differentials with current flowing through the circuit. With high amperage even same resistance will generate voltage differentials that would make it easy to identify problems.

\$\endgroup\$
1
\$\begingroup\$

Best way to test? That is too open-ended a question. We need to know what are your targets or requirements. Does it need to work over a wide range of temperatures? What about a wide range of loads? Is it possible for loads to suddenly become disconnected while in use? What about inputs? Inductive loads (we assume h-bridge = motor control), resistive loads, capacitive loads?

So, identify the regions of operation, create a matrix of all the different cases, and test representative samples of each case - with multiple different h-bridges so you get a feel for how slight differences in board turns affect things and not just testing statistical anomalies. That's the "best" way.

\$\endgroup\$
0
\$\begingroup\$

If you are using high frequency with the gate of the FETS keep an eye on induction noise too, this kind of noise can malfunction any electronic device, specially you are saying you are going up to 70 A

\$\endgroup\$
0
\$\begingroup\$

A power supply can be tested with a dummy load such as a high-power resistor or globe. Most power supplies do not deliver full rated output on a continuous basis. The demand is constantly rising and falling. A dummy load will put a constant demand on the supply and you have to be careful not to over-heat anything. Once you have decided on the value of the dummy load, you will need a power resistor or a globe. There is one major problem with a globe. It requires about 6 times the normal current to get it to start to glow. This is because the filament is cold and its resistance is only one-sixth the operating resistance. Many power supplies will not be able to deliver this current and a globe will not illuminate. You may think the power supply is faulty - so don't choose a globe.

\$\endgroup\$
1
  • \$\begingroup\$ This doesn't seem to address the main part of the question because the OP already seems to be set on using power resistors, although I guess the advice to not use a bulb may be somewhat useful for anyone else considering going that way. \$\endgroup\$
    – PeterJ
    Jan 26 '14 at 23:28
0
\$\begingroup\$

At the risk of being pedantic, if it "should" handle 70A @14V, then you need to get to 70A @14V, not just what's convenient at 12V... Probably more than 70A at 14V, in my book. But the general issues are, - what tests will establish that the design is successful / works / passes? - what is the smallest number of actual test points required? - what's the optimum strategy to work up from small-signal continuity to full load?

So, for starters, how many PCBs did you buy? Will any anticipated failures destroy the PCB? Do you need the PCB to test the components? What is the frequency range of the power passing through? What are acceptable distortions at the output? What is the frequency range of the control being applied? What is the intended relationship between input and output? What is the range of load (0 to 100%? 10% to 100%?)? What is the parallel capacitance range for the load? What is the series inductance for the load? Are there special connectors? Remote sensing of controlled voltage, current, both, at the load?

Simple case, the input is DC, the control is a bit in a register on the board or a voltage threshold on an input to the board, no specific series inductance or parallel capacitance. the 'object' is a switch that turns on and off. The intended load is a power resistor connected with short, low resistance, path.

Complex case? If its intended as a motor controller, then you'll need to control a motor. You might want more than one size (<35A/>35A seems like a big step) and various loads on the motor.

I would define the actual worse case that must be physically demonstrated, which is likely to be somewhat in excess of 70A at 14V, and define the baby steps that lead up to it. 14V isn't much of a big deal, 70A is surely a big deal, so I imagine getting to full operating voltage (and even a bit more) will be fairly easy. But getting to the full operating current (and even a bit more) will require some iteration.

You'll want the unit under test to be firmly attached, over something fireproof. You'll want a fuse (you pick, slow blo? Fast blow?) in series with the supply, to your module, at least. You'll want a fast way to kill the power to your module, a one-switch turn-off on mains power to the supplie(s) driving your unit, a one switch disconnect of the supply TO your unit would be even better.

For my aggressively simplistic case, 5V with trivial current, 5V with 1A, 14V at trivial current, 14V at 1A, A confidence-inspiring over-voltage (14.5V?) at 1A, then double the current 2,4,8,16,32,64A, And 70A. AND the confidence inspiring over-current (72A?) at the over-voltage. Probably want to let it 'soak' at each of the current test points, certainly at the larger ones.

I'd be inclined to have some VOMs hanging on it to show voltage drop across your module, voltage drop across the load, an AmpClamp for current to the load, at least one thermometer on your switch, and probably you want to 'fly' a thermometer probe around to check each output transistor, each output resistor, and put an oscilloscope on the output too. Fly a scope probe to the input, the supply and ground of the module, every connection to the load. Selected points in your feedback loop, if any, or control point(s) that go from whatever your small signal input is to the monster output.

You'll be best off if you make a data collection form and actually fill it out at a few points before full scale and full over-value. I'd make paper copies, pre-filled-out for the test points you've picked for them, but keep some blanks around too. Wanna go paperless? Open an editor with autosave, or better, a spread sheetwith autosave, and have rows or columns for each data item. More than a screen full is too many.

That's a good start. Don't be too elaborate in building up, but don't change more than one condition at a time. I've listed 13 test points above, many would argue that's more than enough for 'bring up' but scarcely enough to prove it works. You have to elaborate on what 'prove it works' means, but turning on and off at 60Hz for a weekend isn't a bad start... Or slow that down to get to full thermal equalibrium at each on/off, or make a mix... you can figure it out. Real world reliability is full rated operating conditions for years...

Please report back your successes!

In my experience, acceptance tests are complete when resuts can be predicted with accuracy and verfied in detail at any selected point. Then you can stop doing bit walks and the like. And change to more complex trials.

\$\endgroup\$

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.