I'm popping MOSFETS....linear derating factor involved?

Thanks Tony!

I wasn't able to figure out how the 208mV figure was derived but I think I understand enough to add the circuit to my breadboard of the active load and play around with it. I'm pretty sure I have a LM339 and/or LM311 around here somewhere.

I'll let you know how it goes.

John

-- remove SPAMMENOT for e-mail responses --

I've got some 11DQ05 Schottky's I can stick in, thanks.

John

-- remove SPAMMENOT for e-mail responses --

[snip]

Abs Max opamp +ve input voltage is when the pot is right at the top and is..... 1.25*R4/(R2+R4), where R2= 49.9k and R4= 10k.

--
Tony Williams.

Ack...of course. And I KNEW that my max was around 0.2V, I designed that in! Amazing how I can zero in on something like "208mV" and lose the big picture.

Thanks Tony, John

-- remove SPAMMENOT for e-mail responses --

That's because there's no such thing as a step, ever.

John

I set up an inverting comparator with hysteresis and it works great...very cool!

I was thinking that instead of biasing the op-amp inputs with the output of the comparator that I could essentially switch Vref (to the servo loops) on/off by using the output of the comparator as the source for the voltage reference. Since the output of the comparator is a lot higher than 1.235V reference, I'd still get a nice steady reference voltage when the comparator is high. And the current draw from the comparator would be pretty low too.

Not sure if this will work but I'll give it a try. I like the idea though as it would reduce my component count a bunch as I scale the active load up in size/power (at least 12 FETs, each with its own servo loop).

John

-- remove SPAMMENOT for e-mail responses --

Or simply power the reference from the load.

RL

Hmm...almost too good an idea to even consider. :-)

Trouble is, the load can go down to 0.5V and I haven't seen any 1%, or better, voltage references that can work with an input voltage that low. I only need 0.2V out so there is "room" for a reference to work there though.

Time for another web search!

John

-- remove SPAMMENOT for e-mail responses --

Clamping or switching the Vref (even down to 0v) is not guaranteed to turn all outputs OFF.

The LM324 has a typical input offset voltage of 3mV. Even if Vref is an actual 0v there could still be some opamps whose offset is in the wrong direction and therefore 'think' they are receiving a 3mV input.

3mV is equivalent to a 300mA output current demand. Sounds tiny compared to 12.5A, but the MOSFET has no drain supply, so it has no way of sinking even 300mA through the current shunt. With no feedback voltage from the current shunt the opamp will *still* turn the MOSFET fully ON.

Whatever scheme is used must guarantee that the -ve input of the opamp is at least 3(5)mV more positive than the +ve input. This means either driving the Vref negative, or spending the components to lift the -ve input.

--
Tony Williams.

Awwwwww, darn....you're right. Just when I found some voltage references with a shutdown pin to drive with the comparator. :-(

I am happy that I actually understood your explanation (which have always been great, it's me that needs to learn a lot more) though so I may just find a way out of this yet. :-)

John

-- remove SPAMMENOT for e-mail responses --

What I thought would be a solution to this was a reference with a shutdown pin that I could drive with the comparator that's monitoring the battery voltage (LOAD+ pin). I could then easily bring the servo loop's Vref to 0V when the battery was disconnected. But Tony Williams brought up a nasty scenario that requires a negative Vref (just a bit though) to prevent.

From Tony Williams:

John

-- remove SPAMMENOT for e-mail responses --

How about having the comparator's output drive the supply pin of the servo loop op-amp? The LT1013's I'm using only draw 0.5mA max so a comparator that can source 6mA would work (I would eventually have 12 FETs, each driven by a LT1013). Or, I could drive a FET to turn the power on/off to the op-amps.

The LT1013's data sheet has an absolute max voltage rating of "equal to the supply voltage" for the inputs though. Since the power is off, I'm not sure if this means that an input can be fried with the 0.2Vref going to it when the power would be off.

LTSpice'ing this scenario (using my original schematic, without the comparator) shows a smooth 1mS ramp up of the gate voltage of the FETs as the op-amp powers up to 12VDC. Not sure if this has any relationship to the real-world scenario of powering up an op-amp with already active inputs though. :-)

May have to wire this one up and scope it out.

John

-- remove SPAMMENOT for e-mail responses --

LOL. I asked for that....

OK, you can always create a ramp that a controller cannot follow. which is a tortured way of saying real controllers have some form of slew-rate limitation.

Similarly, I had a power supply guy tell me the other day that their over-voltage circuit was "instantaneous". When pressed, instantaneous turned into 100us, which is a long time for a junction.

Cheers Terry

I notice that this thing is manually adjusted. So why not add an on/off switch?

You'd still want some kind of effective pull-down to prevent connector surges through load application of dV/dT, through crss. The 1K resistors currently isolate the gates - perhaps a pnp transistor in there somewhere would do it.

RL

I do have an on/off switch but any user of this load would have to remember to turn the unit off before disconnecting the battery and remember to connect the battery before turning it on. That rule will last about a week with the guys around here. :-)

I did try turning the power to the op-amps on/off with the output of the comparator and it works very well when the battery is connected/disconnected cleanly. But, when the battery contacts scrape (and the power goes on for a few uS, then off, repeating this a bunch of times), there are big voltage spikes on the FET gates and huge current surges thru the FETs. I couldn't find any way to slow the system down that would have it essentially debounce the battery connecting/disconnecting.

I really didn't understand this but it sounds like what I was just talking about? That is, connector noise leading to lots of big dV/dT spikes and as the spikes reach the gates the current spikes like crazy too?

I would add a transistor in line the gate to be turned off when the battery was disconnected? The switching would have to be very slow though to provide a debounce function.

Thanks for your help RL! John

-- remove SPAMMENOT for e-mail responses --

Vcc +--------------------------. 1K | | | +------+-------3M-----, | ,---------- LOAD+ | | | | LT1013 | 1.25V POT --+---+1K-+ G +------+-, | +--|-/ | |

Thanks RL! I wasn't able to find the Spice model for the BZX84C2V7L zener but did find a model for a BZX84C3V0LT1. Looked close enough (hopefully).

Running the sim gave me a huge voltage spike from the "battery" at about 189.4mS. The voltage source indicates a linear rise from 0V to

12V starting at 200mS so I'm a bit confused about the timing. I'm probably missing something somewhere though as I'm a SPICE newbie.

Using IRF1405's I get a single 128A spike (with a tiny bit of ringing) thru the FETs when the battery goes high.

Using IRFP2907's I get a single 145A (or so) spike but then the circuit oscillates at 0-14A thru each FET for the rest of the sim.

The transconductance of the 2907's is a lot higher than the 1405's so that might explain the current oscillations. The gate voltages do oscillate a bit for the sim using the 1405's but I guess they're not sensitive enough for that to be a problem?

Gotta stomp that initial spike too but I can play with the model to see how that might be done.

John

-- remove SPAMMENOT for e-mail responses --

I've repeatedly said it here before on s.e.d., but I'll say it again, spice power MOSFET models fail to account for operation as linear circuit elements, where you're near or completely in the "subthreshold" region, and they can therefore completely goof up circuit modeling containing them. Measure your MOSFET's transconductance curves (or measure Id vs Vgs) and compare this to the spice model. Use Vds = 2V. At low currents you can use continuous measurements - but use pulsed measurements at higher currents, where the power dissipation exceeds say one watt.

After you see just how horrible the models are, you can use Google to learn my solution, and that of others.

--
 Thanks,
    - Win

Or learn how to fit data to Level=7 models ;-)

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.