CC load - interesting solution current sharing between mosfets

[mawyatt]

The high open-loop gain of the op-amp absorbs the differences between the active current handling devices and exhibits a low offset over temperature to maintain a balanced current sharing.

Here's an extreme example of this, all 4 current sinking sections have different active components!!

1st section has a Darlington composed of a 2N3904/2N3055

2nd section is 2N4401/2N3055

3rd section is IRF510 NMOS

4th section is IRF530 NMOS

Note how the op-amp absorbs the differences in active current sinking devices, and how all these have equal share of the load current. If you try this with discrete current sensing transistors the results won't be good!!

Best

[LaserSteve]

How to optimize for slew rate?  I'd like to hit 30,000 small jump changes for second.....

Steve

[mawyatt]

That would likely favor the npn output devices (Darlington), and require a faster op-amp with optimized compensation. What we've shown was based upon no analysis, which would be required in any actual use with load and supply extremes as well a device variations.

The sense resistors, with a reasonable offset op-amp, are probably the biggest static error source. Also how well would the discrete transistor version work with low currents?

In engineering everything is a trade off, and the op-amp version seems to be a good candidate for most uses.

Anyway, just another option for the OP to consider

Best

[Yansi]

The high open-loop gain of the op-amp absorbs the differences between the active current handling devices and exhibits a low offset over temperature to maintain a balanced current sharing.

Here's an extreme example of this, all 4 current sinking sections have different active components!!

1st section has a Darlington composed of a 2N3904/2N3055

2nd section is 2N4401/2N3055

3rd section is IRF510 NMOS

4th section is IRF530 NMOS

Note how the op-amp absorbs the differences in active current sinking devices, and how all these have equal share of the load current. If you try this with discrete current sensing transistors the results won't be good!!

Best

Everybody (hopefully) knows, that each individual current regulator will do its job and each section will have the same current.  Now show us how to add a single precision shunt for sum of all currents and a main current controller, that will make precision adjustment of total current from the precision shunt.  That is where the real fun begins.

Not to mention, this thread is about quite different solution for this problem, than individual opamps. Clean opamp solution would likely make a good thread by itself, with all its quirks.

That’s far too complicated. Use cheap TIP142 Darlingtons with an emitter resistor.
The MOSFETs are a poor choice from the outset: they’re not designed for continous DC current, lack a proper high-power package and need a high V_GS and are obsoleted.
More importantly, for a robust circuit you need a fast current limiter. The emitter resistors also act as a current limiter when a BC846 is added.

Whats complicated? I do not think that tuning a bank of opamp regulators with inner loops and outer loop with another opamp is any simpler.
Yeah, mosfets are poor choice, so thats why most if not all commercial loads are built with them today. High Vgs is a problem? Those ~4V are problem why?

No, BJTs are the problem, because of secondary breakdown. Any ideas how many BJTs you'd need to build an average 300W 150V capable load? I'd guess plenty more than with FETs.
Lack proper high power package - what the hell? Its the same TO220/TO247 for either. Whats the trouble there?

That would likely favor the npn output devices (Darlington), and require a faster op-amp with optimized compensation. What we've shown was based upon no analysis, which would be required in any actual use with load and supply extremes as well a device variations.

The sense resistors, with a reasonable offset op-amp, are probably the biggest static error source. Also how well would the discrete transistor version work with low currents?

In engineering everything is a trade off, and the op-amp version seems to be a good candidate for most uses.

Anyway, just another option for the OP to consider

Best

What problem you expect at low current from the proposed BJT based current balancer circuit? If the current is not well balanced at low current, there is zero risk, as power dissipation is low. Absolute current accuracy is still dictated by the single opamps offset voltage, drift and the shunt resistor (with its all non-idealities).

[Yansi]

How to optimize for slew rate?  I'd like to hit 30,000 small jump changes for second.....

Steve

For that you'd need to optimize even the connections between your load and the source supplying it. Depending on the expected current step and source voltage, you may need to have the connection with very low inductance, almost no leads at all. Have a look at AN133 from Jim Williams. https://www.analog.com/media/en/technical-documentation/application-notes/an133f.pdf

Yes, it is a purpose built load, but some principles still may be used, depending on your needs. But 30 kHz load steps a second, that might indeed need some serious bandwidth.

[Jay_Diddy_B]

Yansi,

Check this thread on the forum:

https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/

I went through the design of a dynamic load. I used one op-amp per MOSFET.



Regards,

Jay_Diddy_B

[moffy]

As much as I like the idea of the 'differential quad' transistor configuration, it has the same problem as darlingtons in that there is a base current error, in Yansi's case the quads subtract base current from the measurement. I would love to see it working, but the individual opamp per driver is a more practical solution.

[mawyatt]

Now show us how to add a single precision shunt for sum of all currents and a main current controller, that will make precision adjustment of total current from the precision shunt.  That is where the real fun begins.

Well here you go!!

Just add an extra op-amp that can sense the total load current thru 1 precision sensor resistor Rs and place the other op-amp based sources inside this "loop" that have individual non-precision current sense resistors. Of course you'll need the usual loop compensation and such.

Since you don't need absolute precision balancing with the scheme, as all the "cells" are inside the main precision loop with overall load precision current sensing resistor, you can get by with just standard components. Only the precision sense resistor and main op-amp determine the overall current sinking precision

Anyway, we're sure there are similar schemes that can be employed to solve this

A possible interesting derivative of this concept would make one or more of the current sink "cells" optimized for fast edge speed, while the others are slower. The fast "cells" would respond quickly to the input pulse while the slower "cells" would "catch up" and help share the load and dissipation.

Best

[PCB.Wiz]

Check this thread on the forum:
https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/
I went through the design of a dynamic load. I used one op-amp per MOSFET.

A minor detail is that needs GND-capable opamps, and the TL074 is not GND CMVR, it is hi-side tolerant as the new TL074H

[PCB.Wiz]

As much as I like the idea of the 'differential quad' transistor configuration, it has the same problem as darlingtons in that there is a base current error, in Yansi's case the quads subtract base current from the measurement. I would love to see it working, but the individual opamp per driver is a more practical solution.

True, but unlike std BJT, the base current effect here is very small, around 12ppm for 500mA FET current and modest HFE ~ 300, because the correcting transistors only carry the offset tolerance current.
A larger drawback is the negative supply requirement.
 

[udok]

That’s far too complicated. Use cheap TIP142 Darlingtons with an emitter resistor.
The MOSFETs are a poor choice from the outset: they’re not designed for continous DC current, lack a proper high-power package and need a high V_GS and are obsoleted.
More importantly, for a robust circuit you need a fast current limiter. The emitter resistors also act as a current limiter when a BC846 is added.

Whats complicated? I do not think that tuning a bank of opamp regulators with inner loops and outer loop with another opamp is any simpler.
Yeah, mosfets are poor choice, so thats why most if not all commercial loads are built with them today. High Vgs is a problem? Those ~4V are problem why?

No, BJTs are the problem, because of secondary breakdown. Any ideas how many BJTs you'd need to build an average 300W 150V capable load? I'd guess plenty more than with FETs.
Lack proper high power package - what the hell? Its the same TO220/TO247 for either. Whats the trouble there?

The IRF530 is useable maybe up to 50 Volt (with safety margin due to 2nd breakdown).
Therefore i have assumed less than 50 Volt for the load.

For a 150 Volt / 300 Watt load you would need 9 TO220 or 4 TO3P devices.  Rth is 1.7 for TO200 and 0.8 for TO3P.

In the BJT case the load sharing can be done with simple 0.1 Ohm Resistors and a fast current limiter with a single MPSA42 BJT is trivial.

NPN have 2nd breakdown problems but these and well understood and are documented in the SOA
Modern Mosfets have similar 2nd breakdown problems but the SOA in the datasheet is of no use.

9 TO220 Mosfets with 9 opamps with protection circuit around each opamp are complicated in comparison to the battle tested 4 NPN TO3P circuit with
4 emitter resistors.
Because of gate caps you needs 9 driver transistors too, preferable NPN-PNP pairs for equal rise and fall times.
And you still have no fast current limiter to protect the mosfets in case of short circuit.
Have you done the layout yet?  You will need at least 4 layers.

In addition the opamps could cause nasty local oscillations and will degrade slew rate and impulse response.

But it is true that above 200 Volt Mosfets are the only choice left.

I do not  know commerical loads and can not comment, maybe they copy and paste existing higher voltage designs.

[Jay_Diddy_B]

Snip ..

A minor detail is that needs GND-capable opamps, and the TL074 is not GND CMVR, it is hi-side tolerant as the new TL074H

PCB.Wiz,

The circuit that I showed has +/- power supplies for the op-amps. GND is included in Common Mode Voltage Range, CMVR, for all op-amps.

I built and tested the circuit, it works fine:









Regards,

Jay_Diddy_B

[PCB.Wiz]

I have tried this configuration of the "triferential" pair. Seems it works  too.

The negative supply detail of this bothered me, so here is an emitter injection diff pair, that needs just a positive current feed rail.
It uses matched pair NPNs typically in SOT363 packages.
Has appx 0.65% of current mismatch as it corrects for 460mV of gate skew.
The testing currents are higher here and they are not the same on all FETS, but for power matching that's tolerable.

[PCB.Wiz]

The circuit that I showed has +/- power supplies for the op-amps. GND is included in Common Mode Voltage Range, CMVR, for all op-amps.

Oops, I missed that detail. 
It does not need the negative supply, it can work fine with a gnd tolerate opamp (or RRIO) and a single supply.