Null Detector based on Conrad Hoffman design check

[hozone]

Thanks @iMo

I don't understand, why have I to measure the battery voltage in different position?

[Kleinstein]

Alkaline cells and fresh charged NiMH cell can be a little over 1.6 V. 4 x1 .6 V would be a bit on the high side for the AD8628. I would suggest using diodes in series for the reverse polarity protection. These would reduce the voltage enough to stay below 5.5 V.

[iMo]

..I don't understand, why have I to measure the battery voltage in different position?

It is up to you how you wire the switch, sure, my idea has been to organize the switch positions from "Battery"->100mV->10mV->..10uV in clockwise direction (considering your switch pins numbering goes in reverse)..

[hozone]

Thanks

@iMo
Now I get it!
I've to check the switch I've here.

@Kleinstein
I thought NiMH where 1.2, indeed you are right, they can go up to 1.6V. I've put two SS14 (0.5 drop each), so that even if batteries are 1.6V I will reach 5.4V.
Do you think is better the NiMh cell configuration or the 5V regulator + op amp virtual groundone?

[Kleinstein]

One could also still use a regulator for one side. There is not real need to have the ground exactly in the center. This does however help with the zero trim.
I would prefer the version with the virtual ground and regulator - it is just more stable supply.

[donlisms]

If there is already an input offset adjustment (RV1), couldn't you just disconnect the supply from the Wheatstone bridge and include the voltage due to offset current in the overall offset adjustment?  The bridge is essentially just a resistor across the null detector input in that state, yes?

[Kleinstein]

If there is already an input offset adjustment (RV1), couldn't you just disconnect the supply from the Wheatstone bridge and include the voltage due to offset current in the overall offset adjustment?  The bridge is essentially just a resistor across the null detector input in that state, yes?

That is possible, though it takes the additional step for the offset adjustment for each resistor values separate. With a digital readout this may not be so bad, as one could subtract the offset digitally and thus have a relatively fast zero.

With a separate offset and bias adjustment one could have the trim part stable. Especially with AZ amplifiers the small resitdual offset (a few µV) and also the bias is quite stable. So no need to adjust very often.
For a PCB, I would plan with both. One can still decide later to populate or not to populate.  Higher value resistors for the input divider would be good anyway, at least a factor of 10. 1 M resistors are still available as thin film types with low TC / tolerance.

[hozone]

I was trying to simulate this circuit using LTSpice to understand the changes I have to do for the input bias compensation. I admin I don't yet understand where I have to put compensation resistors 
But I've find something strange, for sure it's my mistake but I don't understand where.

Find attached the .asc for LTspice. and Run simulation
Shouldn't I have 10mV at output when set as 10mV in input?

[iMo]

Your amplification is 1+20000/100 = 201x..
10mV * 201 = 2.010V

[Kleinstein]

The gain setting are only approximate to the round numbers (good enough for analog and digital may have to do some math anyway).

The simulated circuit is still missing part of the divider. So the 10 K or so towards ground.

[hozone]

 

Thanks! 10.09k to ground solves it. Also at 10mV setting, 1mV in is 10mV out.

Now, as for the bias compensation, you mean adding a space on PCB for eventually populating R9 and R10?

[retroware]

You probably want a trimmer in the bias circuit so you can adjust it.  See for example what Keithley did:

http://www.ko4bb.com/manuals/192.31.192.5/Keithley_155_schematic_restored.pdf

To adjust the offset bias current, one needs to be able to measure it.  Keithley takes advantage of the fact that the input resistance of the their meter in the microvolt range is 1M. They measure the voltage across this resistor (input to the meter is open) in order to measure the bias current.  See section 4.2.5 of the K155 manual for the procedure they use. If your input is resistance significantly less than that, you'll have a harder time measuring small bias currents.

[hozone]

So, something like adding a couple of test point between the two R9 and R10 resistors in the ltspice schematic and the space for a trimmer on bottom of each one.

[Kleinstein]

One only needs one large resistor, possibly even more than 100 M. The choice of sign would be more in front of a trimmer. So maybe a 10 K trimmer to ground and some 1 M to either the positive or negative supply. The large resistor would be from the wiper. To reduce the range without changing the trimmer maybe an optional resistor in parallel to the trimmer.

[hozone]

I'm not sure I've understand correctly,
You mean like the LTSpiece attached? (R5 and R9 are the 10k trimmer) R12 is the optional parallel resistor.

[Kleinstein]

That is about the circuit I had in mind, except for the link to short out R9.

[hozone]

Thanks!
Monday I'll put it in KiCad and share it here.

[hozone]

Find attached the schematic.
If it's all ok, I think I'll try to build one. First I'll check the footprint of the rotary switch that I've buy.
Let me know.

[Kleinstein]

The low bat detector still uses quite some power. A simpler version would use the 5 V regulator as a "reference" and than compare something like 45% of the input voltage to the 2.5 V.  For this one could use an OP-amp  (e.g. use MCP6002 together with the virtual ground).

I would still want a footprint for a ceramic capacitor in front of the regulator. Low impedance at the input side is usually helping the regulators.

The capacitors C8 and C9 should be low leakage types - so more like film types that are usually easier to get as THT parts. Beside leakage, also the dielectric absorbtion from X7R or similar capacitors would be an issue. If SMD is really needed this would be C0G types and these would likely need a larger footprint, maybe a little less capacitance. C10 is a little less critical, by I would still prefer a film type here, just to be on the safe side (X7R may need a little more settling time).

The bias compensation may want an extra footprint for an alternative to R23, but to the negative side. The bias can have either sign.
The resistors R3,R4,R5 may want a larger foot-print (at least 0805 ideally even more), so they can better withstand transient voltages.

Ideally the meter would have a way to indicate clipping, e.g. from mains hum. So some kind of comparators to check if the output exceeds some +-2.3 V at any time. A slow DMM may not notice this.

[iMo]

MCHP recommends COUT = 1 μF Ceramic (X7R), CIN = 1 μF Ceramic (X7R) for the 1703..

[hozone]

Thanks,

- added ceramic 1u out of the 5-V LDO, and on input side too
- C8,C9,C10 replaced with Polypropylene Film Capacitor (CBB)
- added a selector for the Vref of Bias Compensation
- changed R4 and R3 to 0805
- added a +-2.1V almost peak detector

Later I'll check the low voltage detector.

SMD has been here for simplicity, I mean manufacturer build and assembly SMD for low cost

[hozone]

Updated schematic and layout attached, still have to check the low battery voltage detector.

[iMo]

R25/26/27 - why not a bit higher values? Like 10x for example?
Hopefully the blinking "Peak" LED will not disturb the equilibrium there.. 
PS: is it possible to make the Peak comparator with a single opamp?

[hozone]

I've review my simulation on falstad.com (attached), with the 1N4148, 0.7V drop almost, maybe it's better to change values.
I've +-2.2V detector changing R26 and R27 to 47, and R25 to 20k.
So it may work better.

[iMo]

Another variant of the Peak Indicator..
With R1 you may set the threshold..