Precision current source single Mosfet or twin transistor setup?

[Overspeed]

Hello

In the aim to built a precision current source 100 mA I have two solutions ( I try to learn )

• one Mosfet as the circuit linked ( from Eevblog forum ) this circuit is a 1mA circuit
  two transistors ( PNP / NPN ) as the circuit linked ( from M. S. Mrozowski design )

This question is a general question on the circuit design so what is the advantage of the both solution

both are voltage driven Op Amp source tested working with proven results.

Purpose is to built a 100 mA circuit for low resistance measurement , I agree than BS170 can be not enough , perhaps a paralleling Mosfet setup can solve the problem or a bigger Mosfet ? Mosfet do not need gate current tuning too.

[Kleinstein]

Both ways can work. Which one works better also depends on how the voltage at the resistor is measured as they have a different common mode voltage for the DUT. Another point could be if one needs extra protection in case there is an external voltage - this could be tricky in the 2nd version.

With 100 mA the leakage current at the MOSFET gate is usually not that critical.
One the other hand 2.5 V would be a bit high for the reference resistor to handle the power.

[PCB.Wiz]

This question is a general question on the circuit design so what is the advantage of the both solution

The second design has the unknown part in the feedback, both terminals floating, and is less ideal for measurement.
The first design is a current sink, not source, but maybe you are ok with +ve voltage to one terminal ?

I’d suggest using a power Pmos, to allow grounded load, and a power spread resistor if you must operate from the 14V shown. Add some clamp diodes, in case the unknown element has some inductance.

[Dr. Frank]

Here's my Precision Current Source recycling project:

https://www.eevblog.com/forum/metrology/recycling-of-precision-current-source-noise-reduction-for-low-burden-shunts/msg1432410/#msg1432410

I use a BS170 for its low bias current with an NPN power transistor in a Darlington configuration, up to 100mA.

Frank

[Overspeed]

Both ways can work. Which one works better also depends on how the voltage at the resistor is measured as they have a different common mode voltage for the DUT. Another point could be if one needs extra protection in case there is an external voltage - this could be tricky in the 2nd version.

With 100 mA the leakage current at the MOSFET gate is usually not that critical.
One the other hand 2.5 V would be a bit high for the reference resistor to handle the power.

Hello
Thanks for your answer
Eevblog YT video 577 1 amp source use a RFP3055 which  is equivalent to the new IRLZ14 I link a screen print with a drain to source resistance at 0.2 Ohm

On the Mosfet circuit 1 mA there is no resistor on the gate , normally there is resistor to control ( limit) the current gate ) I have added one on my PCB in process
Regarding the twin NPN/PNP I link the pdf with the complete schematic , the potentiometer are used because that not matched pair

Regarding the 2.5V in the ref resistor , Vishay propose 8 W foil resistor as the VPR220 with low tempco , could be possible to adjust the ref voltage to match the real resistance value ? VPR from Vishay even with 0.5 tolerance are quite accurate in value .

Regards
OS

[Overspeed]

This question is a general question on the circuit design so what is the advantage of the both solution

The second design has the unknown part in the feedback, both terminals floating, and is less ideal for measurement.
The first design is a current sink, not source, but maybe you are ok with +ve voltage to one terminal ?

I’d suggest using a power Pmos, to allow grounded load, and a power spread resistor if you must operate from the 14V shown. Add some clamp diodes, in case the unknown element has some inductance.

hello

I have linked a complete schematic in PDF in my Kleinstein answer

the designer of this circuit claims impressive performances but unmatched transistor request a lot of adjustements compare to a Mosfet 

Regards
OS

Regards
OS

[Overspeed]

Here's my Precision Current Source recycling project:

https://www.eevblog.com/forum/metrology/recycling-of-precision-current-source-noise-reduction-for-low-burden-shunts/msg1432410/#msg1432410

I use a BS170 for its low bias current with an NPN power transistor in a Darlington configuration, up to 100mA.

Frank

Hello
Thanks for the link , use a small Mosfet to drive a bigger transistor is an elegant solution and I suppose by using a quite big transistor ( 1.35 amp) for a quit low current shall allow a negligible thermal problem .

My goal is to built an affordable circuit to compete with a costly Keithley 6220 current source as I have no budget to purchase a $$$$ Keithley and also I need only a DC 100 mA source

Regards
OS

[Conrad Hoffman]

I've always been partial to the MOSFET design. I first saw it in an early Siliconix MOSFET data book but it might go back further than that.

[trobbins]

I'm sure you are on a better path than what I used, but fwiw:

I used the power fet path for 100mA and 1A stable switched CCS.  My requirements were a little different as I was after a transfer CCS reference, or simple bench top reference, which could operate from a 5V USB battery bank, and use an LM336-2.5 that I have on some spare pcbs, and an opamp I had available.  The current was stable enough for 6-digit meter transfer, but needed measurement and maintaining a constant burdon voltage, especially for the 1A mode, in order to maintain transfer conditions.  I didn't use low tempco resistors, but selected two 10W ceramic block for lowish tempco for the reference R (not as shown in schem below).  I used a bypass resistor for the 1A mode to alleviate the FET temp rise.

Regulation from burdon voltage variation was influenced/limited by the FET on-voltage being in the relatively high-gain but non-linear analog region.  Thermal stability was influenced by various part tempcos, including FET on-voltage.  FET heatsink was live so regulation can be sensitive to external leakage/contact.

[Kleinstein]

If a large current range like with the Keithley6220 is wanted, it would make sense to use largely separate current sources for small and large currents. The regulating transistor (JFET or MOSFET and maybe with an BJT for more current) is the least problem. The resistor switching, ref. voltage and the amplifier for the regulation have different requirements. The higher currents like 100 mA to 1 mA have the keep heating of the resistor in mind and want a relatively small voltage and low voltage noise / drift. The small currents in the low µA range no longer care about the heat, but leakage and current noise become relevant.

A big question is also if one need protection against applied voltage. With the higher currents and high external voltage (e.g. 300 V) the SOA of the transistor can become an issue. Instead of high power parts one can consider a turn off if external voltage is detected.

[trobbins]

Here's my Precision Current Source recycling project:
https://www.eevblog.com/forum/metrology/recycling-of-precision-current-source-noise-reduction-for-low-burden-shunts/msg1432410/#msg1432410
Frank

Frank, what bridge rectifier did you use for B1 and B2 for the dc rails in post #2 of that thread?  The scope image suggested to me that the cause of the transients could relate to diode commutation and leakage inductance in the transformer secondaries, which could be mitigated by using fast soft-recovery diodes and a tuned C-RC snubber across each secondary winding (the 100nF shown may not be optimum).

[Dr. Frank]

Here's my Precision Current Source recycling project:
https://www.eevblog.com/forum/metrology/recycling-of-precision-current-source-noise-reduction-for-low-burden-shunts/msg1432410/#msg1432410
Frank

Frank, what bridge rectifier did you use for B1 and B2 for the dc rails in post #2 of that thread?  The scope image suggested to me that the cause of the transients could relate to diode commutation and leakage inductance in the transformer secondaries, which could be mitigated by using fast soft-recovery diodes and a tuned C-RC snubber across each secondary winding (the 100nF shown may not be optimum).

You might probably be correct .. those are very simple, old bridge rectifier components, but I don't know the type any more.
A snubber network, or 100nF capacitors across these diodes might help to improve the noise further.

I already mitigated the transformer coupling by adding those 470nF  capacitors from mains earth to case GND, but a properly shielded transformer would be better.

As said, I simply re-used all the 35 y/o components.

Frank

[Overspeed]

I'm sure you are on a better path than what I used, but fwiw:

I used the power fet path for 100mA and 1A stable switched CCS.  My requirements were a little different as I was after a transfer CCS reference, or simple bench top reference, which could operate from a 5V USB battery bank, and use an LM336-2.5 that I have on some spare pcbs, and an opamp I had available.  The current was stable enough for 6-digit meter transfer, but needed measurement and maintaining a constant burdon voltage, especially for the 1A mode, in order to maintain transfer conditions.  I didn't use low tempco resistors, but selected two 10W ceramic block for lowish tempco for the reference R (not as shown in schem below).  I used a bypass resistor for the 1A mode to alleviate the FET temp rise.

Regulation from burdon voltage variation was influenced/limited by the FET on-voltage being in the relatively high-gain but non-linear analog region.  Thermal stability was influenced by various part tempcos, including FET on-voltage.  FET heatsink was live so regulation can be sensitive to external leakage/contact.

Hello

Thanks for tour circuit
I have a question : what is the 10K//100K resistor as ratio is 1/10 between 100 mA and 1 Amp

Regards
OS

[Kleinstein]

10K//100K stands for 10 K and 100 K in parallel. Ideally they want 9 K to get a divider to 1/10 of the voltage.

[Zoli]

10K//100K stands for 10 K and 100 K in parallel. Ideally they want 9 K to get a divider to 1/10 of the voltage.

Another trick: parallel two 18kΩ for 9kΩ.

[Kleinstein]

Another way for a 1/10 voltage divider is using 6 equal resistors as 3 in series and 3 in parallel. With some matching this can give a rather good accuracy and low TC even with not so good resistors to start with. There are now also ready made dividers in SOT23 and similar small size.

[iMo]

What "precision" do you actually target with that above circuit?

[Overspeed]

Hello

Question is for who ? in my case I need to run 4 points resistivity measurement so I need to be ''better'' than Keithley 6220 , but as not a programmable  value bu a fixed value better can be reachable .

I also don't want to fall under 10 Volt to reach acceptable power level .

tests / measurements are 10 minutes so i don t need 1 year bomb proof stability too as far I am stable .

I link the Keithley 6220 data table for information 100 mA is not amazing

I have not the budget  for new/calibrated 6220 and I don't need all functions of the 6220 too , just a robust accurate ( stable and precise ) 100 mA source for low to very low resistance material , primary idea was to built a zlymex or a Eevblog current source ( 1 amp) tuned to 100 mA , current source recipe don t look to be NASA rocket science as first is to have a stable and accurate voltage ref as ref54 / ref80 or other with a precision op amp and precision low tempco resistor.

Regards
OS

[Kleinstein]

For a resistance measurement one likely would not need the high settling speed of the 6220 and not much current programming other than maybe on / off / reverse polarity. For a 100 mA range only, the main critical part is the current setting resistor. One would likely want a 4 wire resistor or at least seprate traces directly from the resistor.

Depending on the meter used for the voltage reading, one could also consider the variant with the DUT and ref. resistor directly in series and not on different sides of the FET. This version would allow easier switching between the voltage at the DUT and ref. resistor and also an easy polarity switching. However it kind of needs a separate supply, and a ratiometric measurement gets a bit more tricky.  Depening on the resistance range a true ratiometric measurement may not be needed - very low resistors (e.g. < 10 mohm) may well have other limitations.

[iMo]

Question is for who ?
OS

The Q is for you. You want a precision current source, but except the "100mA" there is none other indication what you actually want (provided this is the Metrology section people wait on some indication). Based on some specific parameters (meaning numbers) people may judge on the actual circuit and parts to be used.

[Overspeed]

For a resistance measurement one likely would not need the high settling speed of the 6220 and not much current programming other than maybe on / off / reverse polarity. For a 100 mA range only, the main critical part is the current setting resistor. One would likely want a 4 wire resistor or at least separate traces directly from the resistor.

Depending on the meter used for the voltage reading, one could also consider the variant with the DUT and ref. resistor directly in series and not on different sides of the FET. This version would allow easier switching between the voltage at the DUT and ref. resistor and also an easy polarity switching. However it kind of needs a separate supply, and a ratiometric measurement gets a bit more tricky.  Depening on the resistance range a true ratiometric measurement may not be needed - very low resistors (e.g. < 10 mohm) may well have other limitations.

Hello
Thanks
I agree on the simple need compared to the possibilities o 6220 .

for the 4 wires resistor I agree that in use on the eevblog 577  amp and also on Zlymex ( eevblog post ) I link the picture of the circuit and the current values.
architecture looks similar , a precision voltage source Zlymez use a AD780 and a VR4 resistor and Eev use a ltc6655 and a VPR221Z with an OPA376

both voltage source are 2.5V

A concern is the resistor heating as the sense resistor handle the power so a direct link wih the voltage and the current .

Regards
OS

[Overspeed]

Question is for who ?
OS

The Q is for you. You want a precision current source, but except the "100mA" there is none other indication what you actually want (provided this is the Metrology section people wait on some indication). Based on some specific parameters (meaning numbers) people may judge on the actual circuit and parts to be used.

Hello
I agree , base on data provided by Zlymex on his 1ma source ( see pic linked ) and the table linked or the 6220

concern is more the current level increase , more difficult to reach tight specs . 

Times Electronic 1021 DC Current Source is advertised for ± 0.02 % of setting ± 0.02 % of range, ± 0.2 μA and 25 ppm/hr stability which sound acceptable

Times Electronic 1024 is accuracy ± 0.02 % of setting + ± 0.005 % of range + ± 0.2 nA

As I measure low resistivity alloy I need 10 mA and 100 mA , problem with 100 mA is the Keithley 6220 is not wonderful on he 100mA range with 0.1 % accuracy and no real value regarding the noise or the stability under load

I need something better than the performance of the 6220 without any other option as that a 100 mA only DC Times Electronic 1021 looks better for my purposes.

I will try the Trobbins circuit with selected resistor and a ref55xx , perhaps set two Mosfet in // can improved the thermal concern of the Mosfet ? in the aim to built a better 1024 time Electronic model

Regards
OS