Is there someting to improve in this design? - Milliohm meter

[Arek_R]

Hi I'm making Milliohm meter basend on the design from "Scullcom Hobby Electronics" youtube channel:





And I was thinking if there's something that I could do better on my schematic, any suggestions?

[spec]

+  Arek_R

Hell, that is a nice piece of engineering: very professional. 

Attached is another approach, which I have had on the back burner for a while. Thought you may be interested. The schematic is hot off the drawing board, and has not been prototyped or tested, so it may have some errors, but it's the principle that I am trying to convey.

[Arek_R]

Thanks
That looks like a EEVblog electronic load xD
It's a good idea as well, but I will stick to my concept as I have all the parts and I like it, was just asking maybe swap one or two components and add something there and there.

[spec]

Let us know how it goes. 

[Gyro]

Here's a Maxim app note (AN106) that shows a possible alternative current source. Deriving the test current from a 1.5V cell would offer a longer life than step-down from a 9V battery. Ref, opamp and mosfet can of course be substituted.

http://www.wentztech.com/radio/resources/Projects/AN106.pdf

[mvs]

was just asking maybe swap one or two components and add something there and there.

ADS1115 has pretty good PGA (differential input, zero drift, low gain error). How about throwing away ina106 and max680?

[Arek_R]

Hmm, I never was thinking about using ADS' internal amplifier, didn't think it would do any good, will put everything on a breadboard and will compare both concepts, but for that I will have to wait for so8 adapter from ebay.
So I use maximum gain of x16(+/-0.256V)?
What about input protection, some schottky diodes maybe?

[mvs]

So I use maximum gain of x16(+/-0.256V)?

With PGA you can implement ranges and switch gain on the fly. Current source will probably go up to 2V, so you can also use 2.048V range.

What about input protection, some schottky diodes maybe?

Input multiplexer has already clamping diodes (10mA).

PS On schematic you have 4.096V version of LT1634, it should be 1.25V.

[MasterT]

In the original design they use panel meter, basically voltmeter they can't adjust, so all this precision 100 mA bla-bla, 4+ resistors in parallel  makes sense. As you introduce uCPU into design, current doesn't have to be precise 100 mA anymore, and can be equal to any reasonable value 10 - 200 mA. Doing re-calibration on start-up by software would allow to correct all nuances.

[Arek_R]

Thanks!
Oh yeah that should be 1.25V version, sorry eagle doesn't have much when it's about such a unusual components, very often I end up using different component just with the same footprint/symbol...

...As you introduce uCPU into design, current doesn't have to be precise 100 mA anymore, and can be equal to any reasonable value 10 - 200 mA. Doing re-calibration on start-up by software would allow to correct all nuances.

So throw away current adjustment 100R pot and do it in software?
That's a good idea, thanks.

But because I'm a complete ultra noob in programming, I'm not sure If I will be able to implement all these great functions, was planning on using couple example arduino codes glued together with wet paper xD

[MasterT]

So throw away current adjustment 100R pot and do it in software?
That's a good idea, thanks.

It's more than kick-off one pot. What I'm trying to say, design concept shown in the video was good in the past century, in 80-th & 90-th.  Precise reference, precise current source, same with voltmeter. It cost a lot. Having uCPU, you don't need all this analog precision, as song as it's not drift much in the minutes time frame. Using only one precise resistor in the circuits and doing comparative measurements by dual channel ADC, ads1115 , ads1232 etc, you can greatly reduce the cost of this project.

[Arek_R]

Yes yes one resistor, that's what I mean :p
What do you mean by "comparative measurements" ?

[MasterT]

This picture explains:

https://www.quora.com/How-can-we-measure-the-value-of-resistance-with-CRO

Measure V and Vm using dual channel adc resistance can be calculated w/o knowing a current, since it zeroed out  of equation.

[spec]

So throw away current adjustment 100R pot and do it in software?
That's a good idea, thanks.

It's more than kick-off one pot. What I'm trying to say, design concept shown in the video was good in the past century, in 80-th & 90-th.  Precise reference, precise current source, same with voltmeter. It cost a lot. Having uCPU, you don't need all this analog precision, as song as it's not drift much in the minutes time frame. Using only one precise resistor in the circuits and doing comparative measurements by dual channel ADC, ads1115 , ads1232 etc, you can greatly reduce the cost of this project.

What a brilliant approach- thanks for posting

[David Hess]

It would be a major design change but I would chop the excitation and synchronously demodulate to remove thermocouple errors.  Essentially this means reversing the test current between positive and negative and doing the same on the sense leads.  Otherwise it is not possible to take advantage of the accuracy (low drift) of even modern bipolar precision amplifiers never mind chopper stabilized amplifiers.

This is commonly done in high end multimeters which provide milliohm resistance measurement.

[Arek_R]

And how do you switch the current without any losses and stability issues of switching components?
It can't be just a H bridge out of N channel mosfets right? xD

[bson]

Get rid of the pot - it's better to calibrate and correct the measurements before readout since you have a µC.  No need for anything adjustable.

[David Hess]

And how do you switch the current without any losses and stability issues of switching components?
It can't be just a H bridge out of N channel mosfets right? xD

Since it is a current being switched, the only losses in switching are limited to bias current errors which are trivial (or nonexistent if the design is clever)  and voltage drops which are irrelevant.  An h-bridge made out of MOSFETs could be used successfully although that might not be my first choice.

[Arek_R]

Actually you're right, 4-20mA exists for a reason :p
So what would be a better way to switch it if not h-bridge?

[unitedatoms]

may be add button for "set zero" routine. and may be 1 more button for good known cal standard ( say some independently measured specimen in range of 10 milliohm). i dont know what is this cpu board includes, i assume it has led and lcd.

also a jumper for firmware to choose between 60 hz vs 50 Hz if you decide to measure with immunity to mains background in integer amount of cycles during acquisition and averaging

[Arek_R]

Well its arduino pro micro(mega32u4) + 1'3 OLED graphic display.
But I also may use graphical "8x2" oled display instead.
And now with the amount of functions I may add a 4 membrane buttons, plenty of them on ebay for dirt cheap.
50/60Hz rejection haha, this primarly weekend project turns into quite advanced device xD
Not much time recently tho :/

[David Hess]

Actually you're right, 4-20mA exists for a reason :p
So what would be a better way to switch it if not h-bridge?

Diodes can be used for current switching with the advantage of lower capacitance and lower leakage.  But an h-bridge is easier to understand and likely good enough.

[Lukas]

Seems like you're all missing one trick: Instead of using a DC current, use AC current and all offset errors are gone. See https://www.analog.com/media/en/technical-documentation/application-notes/an98f.pdf page 10 for an example. Instead of using a synchronous demodulator, digitizing the signal using the MCUs builtin ADC might also be feasible.

[kripton2035]

this one is using an ADS1115
http://circuitcellar.com/cc-blog/build-an-accurate-milliohm-meter/