Current Source and DG408 On-Resistance Problem

[Bob McCloy]

I'm working on a project designing a multi-meter and I am currently building the section to measure resistances. I am using a constant current source design from the Art of Electronics, Third Edition, Page 229. I want to be able to measure resistances from 10Ω to 1MΩ.

I have attached the schematic of my current design. The idea is that the op-amp locks the voltage drop across all the 5 top resistors at 1V and then swapping in one resistor at a time gives a known current through the transistor and down through the load. To measure a resistance I start with the assumption that the load is very large and work down swapping in the next value of resistor if it is within the range of the next current value. Each value of current has a maximum unknown load it can provide a measurement for before the voltage drop across the transistor is too small and the transistor goes into saturation.

The full multi-meter is designed around a micro-controller that will be used to measure the values and display the result. I want to be able to have the micro-controller swap between the resistors automatically so it can auto-range and get a more accurate measurement. I thought about using something like the DG408 Multiplexer. https://www.vishay.com/docs/70062/dg408.pdf and this is where I run into a problem.

My question is, the DG408 has a certain On-Resistance that can be up to 100Ω and also varies with temperature this would seriously effect the current when on the lowest resistance setting of 100Ω and the whole design is based on knowing the current for each of the 5 resistors. Is there any change to my circuit that could negate the effect of the On-Resistance of the DG408. Or any other suggestions on improvements to the design?

Thanks

[RES]

https://www.edn.com/design/test-and-measurement/4388713/Measure-resistance-with-a-microcontroller

[OM222O]

you are almost correct but a few simple changes can fix your issue.
I have built a very similar circuit with current ranges of 10uA to 1A! works fine to 100mA actually but the 1A range is kind of iffy (I got 700 and a change mA output, not sure what's wrong but trying different parts to see if that will fix the issue).
also replace the pot with a voltage reference. if you desperately want 1V, go for it, but I couldn't find any with decent temperature and long term drift for cheap. If not, use a 1.024V reference, you will need to measure it and do the math in your MCU however.

here is the full thread for my project, I think this is exactly what you want:
https://www.eevblog.com/forum/projects/feedback-on-milliohm-meter-v2-0/msg2201538/#msg2201538

schematic:


Edit: I had better luck using the MCP6002 instead of the TLV9002. you can also include a 100ohm to 1k resistor on the output of the op amp to reduce oscillations, but you must use a darlington pair for higher value resistors.
The parts selection is not final, I made an evaluation board to test different darlington pairs, op amps and voltage references, but the project is paused due to the fact that I have exams at the moment, so if you can wait, I will finish my exams in may. we can work on designing a multi meter together. drop me a message if you are interested.

[duak]

You can improve this circuit to make it less sensitive to the analog switch resistances with a Kelvin connection.  This uses two analog switches for each resistor;  one handles the load current and the other handles the sensed voltage. This means that the voltage across the selected load current switch is not fed back to the opamp.  Your circuit can be modified with the following (hopefully my description is clear)

 1. the range selection switches are moved to the right of the resistors
 2. a second set of selection switches connect the inverting input of the opamp to the right sides of the resistors
 3. the switch logic is set up so that the switches for each resistor are enabled at the same time

Very few DMMs use this circuit to generate an accurate constant current for various reasons.  Search for "DMM ohms converter"

Cheers,

[OM222O]

However, you can improve this circuit to make it less sensitive to the analog switch resistances with a Kelvin connection.  This uses two analog switches for each resistor;  one handles the load current and the other handles the sensed voltage. This means that the voltage across the selected load current switch is not fed back to the opamp.  Your circuit can be modified with the following (hopefully my description is clear)

 1. the range selection switches are moved to the right of the resistors
 2. a second set of selection switches connect the inverting input of the opamp to the right sides of the resistors
 3. the switch logic is set up so that the switches for each resistor are enabled at the same time

Very few DMMs use this circuit to generate an accurate constant current for various reasons.  Search for "DMM ohms converter"

Cheers,

using my approach the unused resistors are a part of the negative feedback for the op amp, which has a tiny input bias current. let's say a worst case scenario:1uA input current. that is about 0.1v drop across the resistor chain. I know this is terrible for a precision application, but you also have to consider 1uA is huge! the input current for any half decent amplifier is in the order of nA if not pA! so several orders of magnitude better performance. the error is basically negligible.

[Bob McCloy]

Cheers for the help, this has given me a lot to think about and look over. I very much appreciate it!

[Dr. Frank]

As you're designing that CC circuit for a DMM, you need some precision, maybe.
So your copying from AoE contains several flaws.

At first, the TL071 is not suitable in terms of 'Common Mode Input Voltage Range', see also hints on AoE.

The inputs have to deal with +11V at +12V supply, and that is too close, as the difference has to to be min. 4V, you can't take the typical values.
Either another type is required (see AoE), which can handle Vcc /GND inputs, or you supply the TL071 on >=+15V, and the source from a separate +12V.

Secondly, if you want to have a precision current source, you have to eliminate the base current of the output darlington, by using a FET, again see AoE, Fig. 4-13, for a low bias darlington solution. 10mA may be realized with one FET only.

Especially when you use a FET switch like the DG408, you must use 4 point Kelvin connections, i.e. separating drive current and voltage sensing.
Especially for the higher currents, that's also required. Please also take care about the required / allowed supply / input /output voltages of that IC.

Another hint.. the topology of circuit 4-12 B in AoE is nearly identical to the Ohm current source of the famous HP34401A DMM, its circuits being available online.

I propose, that you download the service manual and copy/adapt the topology of at circuit, instead. There you also find good explanations, how the circuit works.

I copied that diagram for you, and you directly see the proper implementation of FET switches in a 4K configuration. On the high side of the reference resistors, that's also accomplished by sensing from the left, and feeding the current from the right, by the 5V zener diode, which also reduces that input swing voltage of the opamp

Frank

[duak]

One other thing is that the constant current is referenced to the +12 V supply and not 0 V or common.  hp implemented a voltage mirror with U201A  in the 34401A DMM to utilize an existing reference voltage.

Dr. Frank pointed out the operational limits of the opamp.  This circuit could be reworked by adding an 8 V regulator that supplies the current range resistors.  The voltage mirror would also be referenced to this voltage.  The opamp itself would still have VCC connected to +12 V.

Given the current values, the rightmost transistor will have to dissipate quite a bit of power on the lowest resistance range.

OM2220 is right about the range switching - there is a simpliled arrangement based on the Ayrton shunt.   It does require different resistor values because they are switched into the current path sequentialy and have to sum to the correct value for that range.

Cheers,

[Kleinstein]

The series connection of the reference resistors works well for the higher current. It does have limitations for low currents: one is the added noise and bias sensitivity due to the high resistance in series to the OPs input. The other is from the leakage current, especially if a MUX chip is used with all the switches same resistance. So the series connection is good for a simple low grade case without a very small current (like < 1 µA).

It depends on the reference source, whether the simple divider like from the OP or the current mirror version for the reference voltage is better. The divider version needs a relatively high voltage to start with - the current mirror is more like happy if starting from some 3-8 V, but not much more.

For a simple solution one can combine 2 or even 3 different voltages with the reference resistors, so less switching would be needed.  The HP34401 circuit also uses this idea. Especially the lowest current can profit from a smaller voltage drop on the reference resistor, as more voltage is available for the external resistor. The resistor would also be smaller and less critical in the series connection.