Another uCurrent clone, EasyEDA

[analityk]

Good morning everyone.

I want to show you another uC clone, i have publish this project (PCB and schematic) under this link: https://oshwlab.com/Analityk/uCurrentClone

On this occasion, I want to ask you about ground singal on my PCB. I make it with one central point for all GND paths, like you can see in Dave schematic.

It is not my firs PCB ever, i work like "hardware enginer" from few years but in my job we designing PCB with one big ground plane wherever it can be placed, and we connect this with a lots of vias for top and bottom (and sometimes inners layers). But im not sure this methode will be best for uCurrent amplifier.
Firstly i make this project in this "big plane way" but one small bad connection, so i redesigned it. With way shoud be better?

Next question is about something too expensive (but on paper it is realy good) resistors. It they are worth this money? Did someone use it and can confirm TCR or 0,005% tolerance?
Last question is about nA range during OPA189xx can "eat" up to 300pA (worst case is 1 nA) on it noninverting input. I cant imagine (or calculate) how much influence does this have on the measurement result.

Links to resistors i want:
https://www.mouser.pl/ProductDetail/Vishay-Precision-Group-Foil-Resistors/Y145310K0000V9L?qs=adduyb8LSTE261AZJo232w%3D%3D

https://www.mouser.pl/ProductDetail/Vishay-Precision-Group-Foil-Resistors/Y174510R0000B9R?qs=DTo5u7NUXXFxcSg8z45igw%3D%3D

https://www.mouser.pl/ProductDetail/Vishay-Precision-Group-Foil-Resistors/Y14870R01000B9R?qs=Ydpp5fRLlSG5LinmdV7tcA%3D%3D

expecialy this two:
https://www.mouser.pl/ProductDetail/Vishay-Precision-Group-Foil-Resistors/Y00071K00000V0L?qs=ANp4oyDELkFyH31ilT3PdQ%3D%3D

https://www.mouser.pl/ProductDetail/Vishay-Precision-Group-Foil-Resistors/Y00079K00000T0L?qs=Y1RJEkmoluCSlZvoDK8fAA%3D%3D


Thanks for your time,
greetings,
Analityk

PS: i have never be good in english like i want, sorry for my disbraining.

[Kleinstein]

The solid ground plane has an advantage when it comes to EMI and the higher frequency performance. It can also help keeping temperature gradients small.

For most uses one would not need such high precision. At least for the gain setting resistors there are ready made resistor pairs to set a gain of 10 or 100.
For a gain of 10 there is also the option to use an array of 6 equal resistors like in a Hamond divider. The shunt resistors can be the more critical ones and there is not much need to have the gain much more accurate than the shunts. With rather expensive resisistors one may consider adding protection against overcurrent (e.g. parallel diodes).

One can get away with 2 less critical resistors, when the gain is set to x 100 in a single stage of a compound amplifier.

For low currents (e.g. < 1mA) the system with a shunt and voltage amplifier is not the only choice. A transimpedance amplifier gets practical and can have better performance (even smaller burden and lower noise). For the smallest currents one would anyway prefer an OP different from the OPA189. The bias current would add some offset to the lowest range. There is also additional current noise, though the OP189 is still OK with 10 K, but no longer with 100 K or more.

The low current part may be more or less separate from the rest.

The virtual ground part does not need an accurate OP: an LM358 is perfectly good enough there.

[analityk]

Hello,

I've made some calculuation. In scope mA we get 1V per mA. In +/- 15V of power supply i can deliver max 15mA to 10R resistor, it rise 2.5mW losing power. On PCB it is negligible, i think. Opamps in this case take 4mA x 30V = 120mW. Mayby path from shunt resistor to opamp is most sensitive path on board, due to thermoelectric effect and big power (GND/VGND) plane on entire PCB is best options.

Input bias current is negligible. 300pA on 10R can produce 3nV on input voltage. If someone have voltmeter can handle it in 10V range should show it here.

Where is accuracy limit? Ofc my DMM, but from this circuit? I have had this question for a long time.

In my workplace we always use ground plane (and if it is possible - power plane) and it have good results in EMC terms.

I have this opamp for virtual ground because it can handle voltage not smaller than opa189. It was cheapest ~40V opamp in tme.eu

I have measure output offset voltage in first wersion but best DMM in my workplace was 50000 0.03% and have 10uV resolution. So it show me 0.07-0.10 mV (one day after soldering).

Thanks for your time.
greetings,
Analityk

[Kleinstein]

One would normally not use a 30 V supply for the circuit, but more like 9 V or maybe even only 3-4 V. The circuit is typically battery powered and thus relatively low voltage. So no real need for a 40 V OP.

The OPA189 needs some 1.3 mA typical (1.7 mA max) so the power for the OP with a 9 V supply would be more like 12 mW (any less for lower voltage).
With a 30 V supply the power dissipation of the OP would indeed be a problematic point and it can be a reason to avoid such a high voltage.

The idea of the whole circuit is to have a low burden voltage in the 10-50 mV range and with the gain of 100 thus a maximum ouput votlage of some 1-5 V.
With a high burden one would consider less gain and more like avoid the high supply.

[nightfire]

Some thoughts on the supply voltage: As I am having some thoughts about tinkering in the near future with some measurement of the duration of burning time in camera flashes, I would prefer some stable supply without much ripple voltage. So batteries are an affordable way to go.

Here we could have several values: maybe a set of 4 AA batteries or NiMh rechargeables like Eneloop, giving approx 4V-6V during their discharge cycle.
Or a 9V-block battery...
Another popular option are Lithium cells, and here we have with a single cell 3V to 4,2 Volts, and 6V to 8,4 Volts for a double cell setup.
(Side note: I am using some old Lithium battteries from old, broken cameras with some adaptor plate to provide me with a steady voltage for some small breadboard projects- nice recycling of stuff instead of throwing them away)

Question: Which input voltages would your project support, and could they be made in a certain wide range?

[Kleinstein]

There is usually no real need for a high voltage. For most DMMs and scopes some 2 V are large enough to get full resolution and one thus rarely needs more output voltage.
The ranges in the µcurrent are spaced 1:1000 and thus need quite some dynamic range and this can be a reason to sometimes want a little more. Tighter spaced shunts (e.g. 1:100 steps) would resuce the need for higher voltage, but needs a little more care with the scales (no simply A/V, mA/V or µA/V).

For the amplifiers choice, there are quite some amplifiers with a 5-6 V maximum supply. So a low voltge in this can allow more / cheaper OP types.
So the obvious choice would be something like 5 V or a little less directly from a battery.

[analityk]

Hello,

i have discover now my PCB in 10k scope do not work properly. I connect 1.5V alkaline battery thru multiple 100k resistors. With 1.6V source (fresh battery) and 400k resistors it should draw 4uA, so on 10k will be 40mV, after opams it should be 4V. My DMM show 1.6V. It look like my opamps get more current than i expected. I have desolder switch and use only connectors and feed opamps directly (as possible) from shunt.
Did i screw up something or my opamps suck? I have not supposed this was leakage current.
I've repleace 10k with 1k and everythink is like i expect (maybe without comas on dmm).

greetings,
Analityk

[Kleinstein]

There is a slight chance that the circuit may oscillate. The virtual ground is not very low impendance and the signal output can couple back via the virtual ground. It may be a good idea to have some capacitance from the virtual ground to the negative supply (that is parallel to D2).

[analityk]

Hello,

i have replace D2 with 100nF C0G, nothing better, next was 4.7uF X7R, nothing, i changed R3 from 270R to 47R and 1k, nothing, next (last) step was add 1nF C0G parallel to R5 and this solve problem (but with high cost).
I think i should make this feedback trace short as possible, and replace MKP with C0G in small package.
But it work fine, i've build circuit in falstad that carrying 660nA and dmm show me 0,66V. Now it is time to redesign sircuit and buy better resistor - but i thing i looking for cheaper ones.

Thanks 

greetings,
Analityk

[Kleinstein]

The layout looks like it has some kind of guard traces around the input that are done wrong. The output of the fist amplifier is an already amplified signal and this capacitance there adds positive feedback.  The "guard" should be more connected to ground or via a resistor (e.g. 10-100 k range) to the inverting input of the OP.

[analityk]

Hello.

Does it make sense if i place copper zone under trace from switch (amplified signal) to first opamp, especially if it will be connected to inverting input of first opamp?

In my old (first) wrong designed PCB i've cut (or better - i've delete it all) trace connecting guards with output opamp and connet it like you mentioned above, i've add two capacitor (2.2uF X7R) like in reference desing in OPA189 datasheet (beetwen virtual ground and two supply rails), i've replace MKP cap (C5) with 0603 33pF C0G one and all this work was not enough. In 10k shunt scope it show me wrong voltages. Again it all was solved by adding 1nF cap parallel to C5. I belive this insablility was caused by long cables or it is magic. I powering this from battery ofc but when i connecting oscilloscope probe (only "ground") all instabilities goes away.

Realy i have not expected this magic here.

greetings,
Analityk

[Kleinstein]

A copper plane under / around the input = signal from the switch makes some sense. However I would prefer it connected to the virtual ground and not the inverting input. It is usually not a good ideal to have much capacitive loading on the inverting input, though it can be OK in combination with C10.

Another point may be the input impedance for the OPA189. The AZ OP amps may want an about equal impedance at the input. So it may be a good idea to have some R C series element parallel to the 10 K shunt. So something like 100 pF and 100 ohms. This slightly slow down the lowest current range, but could be less than the other 1 nF cap as a fix.

[analityk]

Hello,

i have read carefully all your(s) comments and i have redesign pcb a bit deeper, so all comments are wellcome.
As long as 1nF in feedback is not great sollution only this work on first layout. Today i have measured bandwidth in actual setup and it reach to ~240kHz. (sin wave, from 2.2V outpus signal to 1.7V, so bandwidth should be a bit wider).
This is look prommising from view of new desingned PCB.

glhf