Current Reference IC for V & C Reference Standard Box

[t1d]

I am designing a Voltage and Current Standard Reference Box for calibrating hobby-grade test equipment... I have settled on the MAX6350 for the voltage side... Super 0.02% accurate, right out of the box... But, I am not finding an equally great current reference... I mostly see the REF200, but it only has a 0.5% accuracy... The LT3092 is even less accurate... I was hoping for more... Suggestions? Please and thank you.

[David Hess]

Current references usually rely on the voltage reference and a precision resistor.

[graybeard]

Use a zero drift op-amp, a precision low TC resistor, and a fet.  Connect the gate of the fet to the output of the opamp.  Connect the non-invering input of the op-amp to your voltage reference.  Connect thr invering input to the source and the resistor.  Connect the other side of the resitor to the rail connected to the other side of the reference. The current output will be the drain of the fet.  You can use a enhacement or depletion mode mosfet, or a jfet depending on circuit node voltages.  Use a p-channel device if you want the current source to be on the high side, or an n-channel device if you want it on the low side. 

You can also use a npn or pnp bipolar, but there will be a slight error becasue of the base current.  gate --> base   source--> emitter  drain --> collector (output)

If I am not concerned about fine accuarcy I usually use a bipolar.  It is your choice depending on your needs.

Since for loop gain the fet or bipolar is being operated as a source (or emitter) follower, the op amp loop remains stable so long as you uas a slow (<4 MHz) unity gain stable op-amp.

[Vgkid]

 What sort of accuracy are you trying to achieve, bipolar?
Heres a good look.
https://www.eevblog.com/forum/metrology/recycling-of-precision-current-source-noise-reduction-for-low-burden-shunts/

https://translate.googleusercontent.com/translate_c?depth=7&nv=1&rurl=translate.google.com&sl=auto&sp=nmt4&tl=en&u=http://bbs.38hot.net/forum.php%3Fmod%3Dviewthread%26tid%3D17593%26extra%3Dpage%253D9%26mobile%3D2&usg=ALkJrhiXe_Y6jtKzAUMGsriO3sWrwgY0Jw

[t1d]

Current references usually rely on the voltage reference and a precision resistor.

Well, even with my limited skills, that approach had come to mind. Good thinking. I appreciate your input.

[t1d]

Use a zero drift op-amp, a precision low TC resistor, and a fet.  Connect the gate of the fet to the output of the opamp.  Connect the non-invering input of the op-amp to your voltage reference.  Connect thr invering input to the source and the resistor.  Connect the other side of the resitor to the rail connected to the other side of the reference. The current output will be the drain of the fet.  You can use a enhacement or depletion mode mosfet, or a jfet depending on circuit node voltages.  Use a p-channel device if you want the current source to be on the high side, or an n-channel device if you want it on the low side. 

You can also use a npn or pnp bipolar, but there will be a slight error becasue of the base current.  gate --> base   source--> emitter  drain --> collector (output)

If I am not concerned about fine accuarcy I usually use a bipolar.  It is your choice depending on your needs.

Since for loop gain the fet or bipolar is being operated as a source (or emitter) follower, the op amp loop remains stable so long as you uas a slow (<4 MHz) unity gain stable op-amp.

I think this may one of the best replies to any post that I have ever seen. Why? Because of the wonderful detail. Lots of good info. Thank you.

[t1d]

What sort of accuracy are you trying to achieve, bipolar?
Heres a good look.
https://www.eevblog.com/forum/metrology/recycling-of-precision-current-source-noise-reduction-for-low-burden-shunts/

https://translate.googleusercontent.com/translate_c?depth=7&nv=1&rurl=translate.google.com&sl=auto&sp=nmt4&tl=en&u=http://bbs.38hot.net/forum.php%3Fmod%3Dviewthread%26tid%3D17593%26extra%3Dpage%253D9%26mobile%3D2&usg=ALkJrhiXe_Y6jtKzAUMGsriO3sWrwgY0Jw

Thanks, Vgkid! That is a great thread. The second link did not seem to make sense, though. It took me to cell phone ad?

[t1d]

Here is an article I came across, but I have not worked through it, yet. It appears to be a current reference, having better spec's than a REF200, a "Bolt-On" zero thermal  drift feature and intended to be cheap to build.
https://www.electronicdesign.com/technologies/power-sources/article/21801501/bolton-circuit-transistors-create-zerodrift-current-source

I had wondered if there might be a way to use the Bolt-On feature with the MAX6350, but like I said, I haven't dug into it. If you could clue me in on how to do that, it would be a head-start.

Thanks for everyone's participation and help.

[beanflying]

I built a pair of 6350 VRefs as my first ones (3S LiPo powered) and they have held up well against their initial accuracy over the last few years. They did settle a few PPM during the first few hundred hours so give them a good run. https://www.eevblog.com/forum/metrology/5v-precision-voltage-reference/msg1367777/#msg1367777

There was also an App note  (currently lost  ) using a Darlington 2N6286G (sitting in the project list/box  )as a current source. I have a feeling it was in an earlier datasheet for the LM399 Vref and was good for 10A. If it turns up I will add it here.

[tggzzz]

Super 0.02% accurate, right out of the box...

You should also look at the drift with temperature and, especially, time. You might be unpleasantly surprised. Also be aware that the physical construction can have a significant effect on stability.

How much that matters, if at all, depends on your requirements.

There are many threads on this forum with hints and tips and pointers to reference material.

[t1d]

I built a pair of 6350 VRefs as my first ones (3S LiPo powered) and they have held up well against their initial accuracy over the last few years. They did settle a few PPM during the first few hundred hours so give them a good run. https://www.eevblog.com/forum/metrology/5v-precision-voltage-reference/msg1367777/#msg1367777

Nice project and photos.

There was also an App note  (currently lost  ) using a Darlington 2N6286G (sitting in the project list/box  )as a current source. I have a feeling it was in an earlier datasheet for the LM399 Vref and was good for 10A. If it turns up I will add it here.

If you find it, please do post it.

Not to get too far off track on a voltage reference, but I made a tiny PCB for my MAX6350... Think postage stamp... I socketed the 6350, so I can move it into the combination V & C Reference unit.

[beanflying]

Will do on the Current Source I think it is on and old PC now attached to my Laser so I will have a search. I really should have done PCB's for my 6350's but KISS in the Ozzie bush works for this level of VRef. Make sure you have a read through Andreas comments too in that thread but bear in mind he is chasing sub PPM Voltnuttery for 'fun'

[t1d]

Here is a link about my VRef board. Updated to include the KiCad files, by email, and a free board, if I have an extra.
https://www.eevblog.com/forum/testgear/questions-about-the-use-of-my-new-max6350-precision-voltage-reference-circuit/msg2995338/#msg2995338

[t1d]

Current references usually rely on the voltage reference and a precision resistor.

I have been looking at this, a bit. It certainly is easy and direct.

I wonder about using a digital pot. I understand that it would only adjust in bracketed steps, which would limit achieving accuracy between the steps. However, if there are a lot of steps, it seems to me that a digital setting would be as good as you could dial in with a trimmer, by hand. Thoughts?

[Vgkid]

Try this link, its in chinese.
http://bbs.38hot.net/forum.php?mod=viewthread&tid=17593&extra=page=9&mobile=2

[t1d]

Try this link, its in chinese.
http://bbs.38hot.net/forum.php?mod=viewthread&tid=17593&extra=page=9&mobile=2

Nope, still a vendor ad for cell phones... Look at the link address = "38hot." That's the name of the vendor site. What should I be seeing?

[oPossum]

Fixed link: http://bbs.38hot.net/forum.php?mod=viewthread&tid=17593

[beanflying]

Good link to that Chinese site. Google translate does an ok job 

And just because Bob Pease is always worth a watch. Off to rummage around in my other PC....

[t1d]

That's an amazing link. Thank you.

I will watch the video. Later today, I hope. Thanks for it.

[David Hess]

Current references usually rely on the voltage reference and a precision resistor.

I have been looking at this, a bit. It certainly is easy and direct.

I wonder about using a digital pot. I understand that it would only adjust in bracketed steps, which would limit achieving accuracy between the steps. However, if there are a lot of steps, it seems to me that a digital setting would be as good as you could dial in with a trimmer, by hand. Thoughts?

Digital pots tend to have poor absolute accuracy and are intended to be used as dividers.

A switched capacitor circuit can generate a precision variable current from a voltage within the limits of the accuracy of the capacitor but fixed resistors will be more accurate and easier to use.

[t1d]

Digital pots tend to have poor absolute accuracy

Do you mean beyond the know issue of not being able to select values that fall within the bracketed amounts? Example 10K Ohms/1024 steps... (I am a v-e-r-y novice coder.)

and are intended to be used as dividers.

I am not doubting you, here, but please point me to some materials that would teach about this effect.

A switched capacitor

I know of variable capacitors, but what is a "switched" capacitor?

circuit can generate a precision variable current from a voltage within the limits of the accuracy of the capacitor but fixed resistors will be more accurate and easier to use.

I know about using multiple resistors of the same value to average/lessen the tolerance error. I am also familiar with using a set value resistor and a trimmer. If I understand correctly, I can get an order of magnitude better accuracy with a trimmer.

It looks like the resistor/K.I.S.S. approach is best, for this project. Thanks for your input.

[David Hess]

Digital pots tend to have poor absolute accuracy

Do you mean beyond the know issue of not being able to select values that fall within the bracketed amounts? Example 10K Ohms/1024 steps... (I am a v-e-r-y novice coder.)

For example the Analog Devices AD5293 is 35 ppm/C when used as a rheostat and and 5 ppm/C when used as a divider.  The absolute tolerance in rheostat mode is 7%.

and are intended to be used as dividers.

I am not doubting you, here, but please point me to some materials that would teach about this effect.

Just check any of the datasheets.  Accuracy is better when used as a divider than as a rheostat simply because a divider depends on identical resistors tracking while a rheostat depends on their absolute value plus the "wiper resistance".

A switched capacitor

I know of variable capacitors, but what is a "switched" capacitor?

If a capacitor is switched between two points, then it transfers charge between them proportional to the difference in voltage, capacitance, and frequency.  So it can act as resistance controlled by the frequency of switching.