Electronic Load parallel Mosfet Design

[Bootalito]

I had an idea that I've been kicking around to parallel mosfets in their linear region for an electronic load I've been designing.  It involves a master op amp that is controlling the overall current with individual slave op amps that can be paralleled as many times (within reason) as needed.  I've gone through multiple design revisions while receiving assistance from people here on this forum.   I very much appreciate it.  Would something like below work?

Assumptions:
1. Power dissapation in the shunts is taken care of.
2. The "slave" ballast resistors/shunts are generic cheap 5W awful tempco resistors.
3. The main shunt (R24) is a good shunt with good tempo and proper power dissapation.
4. The AD8630 will probably be replaced with another TL071 or the like.
5. The exact values of passives still needs to be tuned

The only weird thing is that the current used by the slave op amps will flow through the main shunt on its way back to ground which will technically give me incorrect readings (positive) offset) that I'll have to account for in software somehow I guess.  I'm just looking for an overall.."yeah that should work if you've worked all the details out", or "No that overall designwon't work becuase of X".

[ajb]

A more conventional approach would be to skip the first op amp stage and the 'master' sense resistor and supply your programmed reference to each of the parallel current sources.  Your way does save the cost of multiple potentially expensive precision shunts, but your main shunt has to be rated for the full output current, which will make it more expensive.  It also means it must be a lower value to keep your overall voltage drop the same, necessitating more gain and overall having to deal with smaller signal voltages if that's where you're doing your main sensing.  This is not be such a problem on an electronic load as long as your minimum voltage requirements aren't too close to zero.  Compensating the system will be a little trickier since you have an inner and an outer loop to deal with, but shouldn't be too bad if you're willing to accept the consequence of a necessarily slower overall response.

There's no point in connecting the negative supply of U42/43 to the top of the main shunt.  They can be connected to your regular negative supply.  TL07x is not a good choice here, though--without a negative supply you need an op amp that behaves itself when the inputs are close to the negative rail. 

If you aren't already aware, be very careful in selecting your FETs.  Many of the high power parts on the market have a very small DC SOA, so you can't just grab any old FET just based on its voltage and current ratings and throw it into a linear circuit where you need it dissipate some power.

Oh, and R12 is on the wrong side of R13.

[Kleinstein]

The supply of the Tl071 should be directly from normal -5 V (or similar), not through the shunt. The input current to the FETs and OPs is usually very small an can be neglected. Other than that there would be no more wrong current.

The TL071 is not single supply and may thus need a negative supply.

[T3sl4co1l]

Yeah why wouldn't you return the op-amps to ground/negative, putting them to a shunt just makes everything more awkward.

Why trim offset of the crap channels, when you're controlling the total?

Why not use TLV2372 or other single-supply options (even LM358 would be a thing, with a bias resistor loading the output down, to keep it out of class B operation)?

Tim

[Bootalito]

Yeah why wouldn't you return the op-amps to ground/negative, putting them to a shunt just makes everything more awkward.

Why trim offset of the crap channels, when you're controlling the total?

Why not use TLV2372 or other single-supply options (even LM358 would be a thing, with a bias resistor loading the output down, to keep it out of class B operation)?

Tim

In order for the slave amps to measure the voltage drop across the resistor one must necessarily put the ground where I put it.

The supply of the Tl071 should be directly from normal -5 V (or similar), not through the shunt. The input current to the FETs and OPs is usually very small an can be neglected. Other than that there would be no more wrong current.

The TL071 is not single supply and may thus need a negative supply.

There is really no such thing as an op amp that cannot be used from a single supply.  Dual rail op amp requirements are not really a thing.  Although they may not have the performance requirements required to fufill the design, but that's another issue.

A more conventional approach would be to skip the first op amp stage and the 'master' sense resistor and supply your programmed reference to each of the parallel current sources.  Your way does save the cost of multiple potentially expensive precision shunts, but your main shunt has to be rated for the full output current, which will make it more expensive.  It also means it must be a lower value to keep your overall voltage drop the same, necessitating more gain and overall having to deal with smaller signal voltages if that's where you're doing your main sensing.  This is not be such a problem on an electronic load as long as your minimum voltage requirements aren't too close to zero.  Compensating the system will be a little trickier since you have an inner and an outer loop to deal with, but shouldn't be too bad if you're willing to accept the consequence of a necessarily slower overall response.

There's no point in connecting the negative supply of U42/43 to the top of the main shunt.  They can be connected to your regular negative supply.  TL07x is not a good choice here, though--without a negative supply you need an op amp that behaves itself when the inputs are close to the negative rail. 

If you aren't already aware, be very careful in selecting your FETs.  Many of the high power parts on the market have a very small DC SOA, so you can't just grab any old FET just based on its voltage and current ratings and throw it into a linear circuit where you need it dissipate some power.

Oh, and R12 is on the wrong side of R13.

Thanks..yeah R12 is on the wrong side.  I think I may just go with a more conventional approach.

[Mechatrommer]

There is really no such thing as an op amp that cannot be used from a single supply.  Dual rail op amp requirements are not really a thing.  Although they may not have the performance requirements required to fufill the design, but that's another issue.

yes +5v -5v rail can be the same as +10v gnd rail from opamp point of view but the thing is trying to avoid the opamp operating outside its specification namely its common mode range.. at high current, the inner loop opamp's gnd will rise, in case the outer loop opamp tries to stop them by giving hard gnd output, its well under their cm range.. non behaving opamp will give nasty effect such as phase reversal but i think tl071 is not one of them.

Once you get its stability working (which i believe will be the main issue here due to inner n outer control) , if someone want to copy your design with say 4558, then suddenly it goes nasty such as burnt mosfets or worst, dut.

Edit. furthermore, using non rail-rail opamp will make changing mosfet to power bjt in the design is not an option.

Ymmv..

[T3sl4co1l]

I mean, sure, I'm probably wrong, I mean, I'm just some stupid opinion on the internet.  It's not like I've ever designed a dozen things and had them work first try... 

Tim

[ajb]

non behaving opamp will give nasty effect such as phase reversal but i think tl071 is not one of them.

No, the TL07x and its brethren WILL show phase reversal when the lower Vcm range is violated.  I have a funny* story about TL074s being used in a device that was supposed to monitor a split supply and shut down some associated signals in case of power supply failure.  Guess what happened when the negative supply failed or simply came up slower than the positive side of the supply!

* Funny in hindsight, anyway.  At the time it resulted in an awkward series of conversations with the manufacturer. . .

In order for the slave amps to measure the voltage drop across the resistor one must necessarily put the ground where I put it.

Whether or not the output signal of an op amp is meaningfully referenced to ground depends on if its input and feedback networks are referenced to ground (and how).  The negative rail of the op amp is irrelevant, as long the op amp's supply and common mode voltage ranges are observed. 

In your circuit, the slave circuits just sink more or less current through their sense resistors until the voltage at the top of the sense resistor matches whatever voltage the master op amp is providing.  They don't care if either or both is offset with respect to ground, because the master's feedback loop is only closed through the current they cause to flow through the master sense resistor, and therefore its output is not meaningfully referenced to ground.  The master's output will swing as high or as low as it needs to (or can) to get the master sense resistor voltage to match the master set point.  If the master op amp were configured as a simple fixed gain buffer (with DC feedback directly from its output to its input), then the signal at its output WOULD be referenced to ground and the offset would matter.  But it's not, so it doesn't, and even then the negative rail doesn't need to be anywhere in particular for it to work.

There is really no such thing as an op amp that cannot be used from a single supply.  Dual rail op amp requirements are not really a thing.  Although they may not have the performance requirements required to fufill the design, but that's another issue.

Sure, technically, an op amp doesn't care where its supplies sit with respect to ground.  Some, however, very much do care about where their inputs and outputs are with respect to their supply rails!  That's why op amps are spec'ed with a common mode voltage rating.  Saying that an op amp requires a split supply is short hand for saying that it will behave either poorly or even maliciously if you operate it too close to its negative rail.  The TL07x happens to be one of the latter.  If you won't take our word for it, wire one up as a unity gain buffer, sweep its input, and watch what happens on the output.

[Mechatrommer]

non behaving opamp will give nasty effect such as phase reversal but i think tl071 is not one of them.

No, the TL07x and its brethren WILL show phase reversal when the lower Vcm range is violated.

correct! i was wrong typing from a handphone in the office... attached is TL072 i tested (bought in ebay) sometime ago... ch1 input, ch2 output, ch3 +rail ch4 -ve rail... that design will goes sheety if using rail-rail as its outer loop control... fwiw...