Mosfet's current at low Vds (related to CC dummy load)

[BravoV]

This thread is somehow related with dummy constant current load like Dave did, and while reading lots of mosfet datasheets, my eyes spotted interesting fact that at low vds condition, most common "switching mosfets" when it's VDS is at pretty low say below 1 volt, the ability to conduct current is quite low compared to higher voltage say > 5 volt, even the gate already driven like at 10 volt.

Take the MTP3055V as the example, this exact mosfet is also used by Dave at his constant dummy load.
When VDS at 0.5 volt, the max current is only at around  < 6 Amp, also this chart is where Tj at 25 C, and aware that at higher Tj, this number will be even lower.



Aware of other special type linear mosfet like the one made by IXYS, which has an impressive current capability even at very low vds, the chart below from IXTK200N10L2 shows the capability with Tj at 125 C  . Compared to above cheap MTP3055V, its like night and day.


Assuming I have lots of spare MTP3055V or other common switching mosfets lying around, and don't feel like to make a dent in my pocket to order the IXTK200N10L2 above at $25 a pop 


Now questions :

- If I want to use a common mosfet for disspating at high current say like 10 Amps at 0.5 Volt source, can I use MTP3055V or other common cheap switching mosfet in parallel to solve with this single mosfet limitation ?

- How about the opamp that is driving the gates of parallel mosfets or even the overall circuit ? Does it need major modification ? and how exactly ?


TIA


PS :
Aware of the potential issues when common switching mosfet used in linear mode in dissipating high energy, made a thread for it while ago here -> CLICK, but please ignore this issue for a while in this thread.

[electroguy]

- How about the opamp that is driving the gates of parallel mosfets or even the overall circuit ? Does it need major modification ? and how exactly ?

not sure if it helps you, but read post 24 and 28 in the following thread:
https://www.eevblog.com/forum/beginners/dave%27s-constant-current-load-traps-for-young-players/15/
someone suggested to use multiple op amps (ie, 1 op amp per fet)

[Rufus]

while reading lots of mosfet datasheets, my eyes spotted interesting fact that at low vds condition, most common "switching mosfets" when it's VDS is at pretty low say below 1 volt, the ability to conduct current is quite low compared to higher voltage say > 5 volt, even the gate already driven like at 10 volt.

The drain current for a given drain voltage when the MOSFET is turned on is determined by the drain resistance and specified in data sheets as RDSon. No need to zoom in on parts of graphs.

Max voltage, max current and RDSon are the three 'highlight' parameters for switching MOSFETS.

The MTP3055V has max RDSon of 150m ohm, the IXTK200N10L2 has 11m ohm.

[Short Circuit]

Congratulations, you have rediscovered Ohms law

[ejeffrey]

Yes, you can connect mosfets in parallel, but the best way to do it is to follow the advice of electroguy's link.  Give each mosfet its own sense resistor and its own feedback opamp.  This forces the current to split evenly, spreads the load over three sense resistors, and avoids annoyingly small values of the sense resistor.  You would need 6-8 of the MTP3055V with 0.1 ohm sense resistors.

You should be able to find a cheap MOSFET with a low Rds on.  Just search the digikey/farnell parametric search.  I found a number for under $1, but you will have to look up the data sheets to see if they are suitable.  The IXTK200N10L2 is a beast with a 1040 watt (!) power rating.  You only need 5 watts or so.

[BravoV]

The drain current for a given drain voltage when the MOSFET is turned on is determined by the drain resistance and specified in data sheets as RDSon. No need to zoom in on parts of graphs.

Max voltage, max current and RDSon are the three 'highlight' parameters for switching MOSFETS.

The MTP3055V has max RDSon of 150m ohm, the IXTK200N10L2 has 11m ohm.

Doh .. Rufus, you're absolutely right ? Don't know what I was thinking, this is really embarrassing. 

I guess this is what they called brain fart. 

Yes, you can connect mosfets in parallel, but the best way to do it is to follow the advice of electroguy's link.  Give each mosfet its own sense resistor and its own feedback opamp.  This forces the current to split evenly, spreads the load over three sense resistors, and avoids annoyingly small values of the sense resistor.  You would need 6-8 of the MTP3055V with 0.1 ohm sense resistors.

You should be able to find a cheap MOSFET with a low Rds on.  Just search the digikey/farnell parametric search.  I found a number for under $1, but you will have to look up the data sheets to see if they are suitable.  The IXTK200N10L2 is a beast with a 1040 watt (!) power rating.  You only need 5 watts or so.

In parallel while each mosfet has it's own op-amp, about the sense resistor, what if only using single one instead of one for each mosfet ? Its just I happened to have a nice low ohm precision resistor from Vishay.

[ejeffrey]

I am not 100% sure what you mean, but if you have separate opamps deriving feedback from one load resistor you have a defective circuit.  Due to opamp input offset errors it will  send all the current through one FET. If you can only use one sense resistor just put the FETs in parallel with a single opamp.  You will have to be more careful with thermal management and make sure you stay within the SOA even in the worst case current split.

[BravoV]

I am not 100% sure what you mean, but if you have separate opamps deriving feedback from one load resistor you have a defective circuit.  Due to opamp input offset errors it will  send all the current through one FET. If you can only use one sense resistor just put the FETs in parallel with a single opamp.  You will have to be more careful with thermal management and make sure you stay within the SOA even in the worst case current split.

Will this circuit work ? Or not even close ?



The reason why I'm thinking of using multiple mosfets is because I got tons of this common and identical mosfets lying around and brand new, so by spreading the heat generated across multiple mosfets is better than a single one taking the whole heat load.

I hope this thinking it self is not flaw. 

[amspire]

Close, but it will not work.  It will definitely oscillate. The current will not share over the full voltage range. Parallel mosfets in general do not share current properly.

Take your circuit, and strip it back to one mosfet. Add a resistor - perhaps 10K - from the inverting MOSFET input to the shunt resistor. Add a capacitor - perhaps 1nF - from the inverting opamp input to the output. Add a resistor of at least 10 ohms from the transistor emitter to the gate of the single mosfet.

To add multiple mosfets, repeat the same complete circuit for each mosfet and connect all the CurrentSet inputs together. Also connect all the grounds and DUT connections together so that all the current sinks are in parallel.

Now you have something that can work and that will properly share current in the mosfets

Richard.

[BravoV]

Hey Richard, thanks.

Like this ? Please correct me if there are mistakes. 



Just asking, since each mosfet has it's own driver, does it still need that transistor ? How about driving directly from the op-amp output with 10 Ohm resistor to the gate ?

[electroguy]

Just asking, since each mosfet has it's own driver, does it still need that transistor ? How about driving directly from the op-amp output with 10 Ohm resistor to the gate ?

1. yes you can get rid of the transistors and extra circuitry (op amp can drive the resistor/fet directly). Even with your original circuit driving multiple op amps, you didn't need it.
2. you can't share a single shunt resistor across multiple op amps, you need one per op amp.
3. you have the 1nf caps in the wrong spot. They should be connected from output to - terminal of op amp

Read the following thread, and download the version 3 file in post #12 and have a look at the schematic on the left.
https://www.eevblog.com/forum/beginners/derpy-load-is-this-dummy-load-design-any-good/

[amspire]

Hey Richard, thanks.

Like this ? Please correct me if there are mistakes. 

The 1nF caps have to go to the inverting input - not the shunt side of the 10K resistors.

I would put the 1K resistor back on the transistor emitter and lower its value. Perhaps 100 ohms.

Just asking, since each mosfet has it's own driver, does it still need that transistor ? How about driving directly from the op-amp output with 10 Ohm resistor to the gate ?

The transistors will allow the mosfet to change faster.  It is probably not needed if you are expecting the load you are testing to be mostly fixed voltage. If you want to regulate the current well with changing load voltages, the extra transistors may help. If you take out the transistors, have a 1K resistor from the opamp output to the mosfet gate.

Richard.

[electroguy]

The transistors will allow the mosfet to change faster.

just curious, how? all i can see is they add extra delay, but i could be wrong?

[IanJ]

Hi all,

Just some notes:

Driving a mosfet with an op-amp (without a transitor buffer) then watch out for some problems when you are loading up over 15vdc. I found I had to switch to an LM7322 or LM8272 which is designed with high capacitive loads in mind otherwise the op-amp fails to drive properly.

I'm also using a IXTK46N50L mosfet and am able to load up to 10amps @ 24vdc easily, although I did have to add quite a high value integrator cap (4.7uF) as a quick fix to kill some oscillations on gate.......not to sure why yet though.......

Ian.

[amspire]

The transistors will allow the mosfet to change faster.

just curious, how? all i can see is they add extra delay, but i could be wrong?

Emitter followers, particularly if you reduce the emitter resistor to a few hundreds ohms are very fast. The bandwidth of the emitter follower without the mosfet gate as a load is probably in the 100's of MHz.

The thing that slows the circuit down is the gate capacitance of the mosfet, and if you look at the specs, you will see this varies hugely. As a rough guide, the lower the current rating of the mosfet, the lower the gate capacitance so for a dummy load circuit, choosing a really low resistance mosfet is only needed if you need the current limiting to work at very low voltages. Otherwise, a mosfet with a channel resistance of over 0.1 ohms is usually better for both capacitances and Safe Operating Area.

The emitter follower will be able to change the voltage faster on the mosfet gate then the opamp can, and it will be able to cope with the effect of the gate-drain capacitance on the gate voltage in the case the load voltage is changing and not static.

But definitely try it with and without the transistor. Feed a 1KHz square wave into the current ref input and look at the waveforms across the shunt resistor with and without the transistor. In many cases, people are happy with just the opamp driving the gate directly.

Richard.

[electroguy]

good points.
thanks!

[BravoV]

The 1nF caps have to go to the inverting input - not the shunt side of the 10K resistors.

I would put the 1K resistor back on the transistor emitter and lower its value. Perhaps 100 ohms.

Like this ? As eletroguy mentioned, will this work if only using single sense resistor instead of spread them at each mosfet ?

The transistors will allow the mosfet to change faster.  It is probably not needed if you are expecting the load you are testing to be mostly fixed voltage. If you want to regulate the current well with changing load voltages, the extra transistors may help. If you take out the transistors, have a 1K resistor from the opamp output to the mosfet gate.

Richard.

Fine, this transistor cost almost nothing anyway.

[amspire]

No, you have to have one sense resistor per mosfet. It is the only way to get currents to share properly. You must have one sense resistor for each opamp. If you look at Dave's recent tear-down of an electronic load, there was a current sense resistor for each mosfet. The single current shunt on the input was for measurement, and not for the current regulation circuits.

The only places each current sink circuit interconnect to the others is the ground, the DUT connection, the CurrentSet connection, and the opamp power supplies. That is all.

If you want to have one current sense resistor, then you have to source a single mosfet and heatsink that can handle the full power that you need. The mosfet has to have a DC SAO (safe operating area) specified in the datasheet that matches the voltage/current ranges that you are after.

Mosfets can be selected to share current reasonably well for a specified voltage, but you have to make sure from the datasheets that for the conditions where you need more then one mosfet operating, the channel resistance will dominate the gate voltage variations to such an extent that the mosfets will share properly.

You have to really understand the datasheets to do this, as the majority of mosfets will just not share properly at all - one mosfet will heat up and take almost all the current. Most mosfets do not have variations in gate voltages specified in a useful way that would enable you to properly choose such parallel mosfets. I would just not even bother.

Richard.

[codeboy2k]

If you look closely at Dave's latest video, the teardown of the BK Precision 8500 Electronic Load, you will see that they did it exactly as has been described here by Richard, and the other links in other posts.

There are 8 IRFP250N mosfets, 8 opamps (TL08x) , 8 RX21-8W 0.05ohm, 5% sense resistors, and although I couldn't count 8 transistors ,I did see 3-4 transistors scattered around the op-amps.   I am sure they are using transistor gate drivers in an emitter follower configuration, I just didn't see them all.

[electroguy]

Also, don't forget to take a step back and look at what you are really trying to do. There are plenty of mosfets out there in the $1 to $5 range that can handle 10A to 20A at 0.5V, so why are you trying to use something that is not suitable? I know you said you have lots of mtp3055 lying around, and want to reduce cost, but analyse what you are really going to save.

For example, how many of these are you going to make? if you are going to make thousands, and you already have thousands of mtp3055, i can see your point.
but if it is just 1 or a few, then why not spend $2 or $3 on a FET that does what you want? trying to use a $1 mtp3055 fet that is not suitable, just because you have it, isn't really going to save you money. You will end up spending more in the long run (because your board needs to be bigger, you need more parts/circuitry, you need more heatsink space, more mounting hardware, more heatsink compound, more shunt resistors etc...)

So, what is your EXACT goal that you are trying to achieve? 0.5V@10A? or 0.5v@20A? Or is it something else?