In that case you need a reasonably accurate current source of some decimal value. Removing the offsets in a purely analogue way is also pretty complex. Having a 4-terminal reference resistor, some pretty unimportant current and a micro to do the math is so much simpler.
But it is not a Vref issue. OTOH, if the current through the reference resistor is producing voltge drop close to the ADC's full range reading, then for a 200mOhm RREF even a 16-bit ADC would provide ~3uOhm resolution, which is sufficient. I think the key point is to measure V(RREF+DUT) instead of V(DUT). The value should then be pretty close to V(RREF), so the non-linearities of the analogue path will not be important for small values of DUT. For larger ones you already have enough voltge drop in order not to care about the last bits.
Chopper instrumentation amps are pretty good these days, but I agree it's generally better to bias the measurement a bit away from the supplies.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510
http://electrooptical.net
http://hobbs-eo.com
I agree, thermals are a problem. I have done these systems using relays to auto zero, and that was using those expensive HP $5k meters as well!. It was for magnet quench detection in another lifetime at the SSC, of which I no longer remember the details... I now stick to things a couple of mm square nowadays....
My main point, was that I rolled my eyes a bit on suggestions of using a discrete A/D. Just use what is easily available.
I'd suspect that a 16 bit uP running at a couple tens of MHz, plus a good quality diff amp and the proper processing/digital lock-in could get to 10s of uOhm accuarcy on sub 1 ohm measurements with the onboard
10 or 12 bit successive-approximation ADC, no fancy sigma-delta or chopper required.
Seems good enough for a not-as-good-as-Tektronics bench instrument to me....
The few high-resolution ADCs I tested recently do stay still - both the AD Sigma-Deltas and the Linear SARs. The best ones have white noise in the bal lpark of some tens of nV/sqrtHz, so you do get >20 bits even at bandwidths well above 10 Hz. That's with shorted inputs or in fully ratiometric mode. The output data rate doesn't affect the noise density - it just determines how much of the bandwidth you throw away. When it comes to 1/f - again, in the best device it's super impressive - the corner is below 0.1 Hz.
For comparison, the internal bandgap Vref of AD7177 has white noise of 200 nV/sqrtHz and terrible 1/f noise. Any non-ratiometric measurement with this Vref is doomed to very sub-optimal performance, considering the otherwise excellent capabilities of the ADC.
Furthermore, with these parts you can completely forget they are Sigma-Delt a. You don't see _any_ of the modulator quantization noise. The modulator r uns at 8 MHz, so the corner frequency of the shaped noise is somewhere in t he tens of kHz, which is above the maximum data rate they offer.
With 20 bucks for an ADC and 20 for the rest (external reference, amps, etc .) you can get quite far these days. The rest is just a standard SPI interf ace that you hook to your favourite MCU.
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510
http://electrooptical.net
https://hobbs-eo.com
AD Sigma-Deltas and the Linear SARs. The best ones have white noise in the ballpark of some tens of nV/sqrtHz, so you do get >20 bits even at bandwid ths well above 10 Hz. That's with shorted inputs or in fully ratiometric mo de. The output data rate doesn't affect the noise density - it just determi nes how much of the bandwidth you throw away. When it comes to 1/f - again, in the best device it's super impressive - the corner is below 0.1 Hz.
AD7177-2, AD7176-2, LTC2378-20 and LTC2380-24
formatting link
I don't have the 7176 in this plot. It had similar white noise as 7177, but higher 1/f and worse linearity. Also, 7176 is capable of higher data rates , so you can actually see the quantization noise in the tens of kHz.
These are all measured with shorted inputs so there's no noise coming from input amplifiers, Vref, etc. The full-scale range is either +-5 V (SDs) or
+-4V (SARs). They all run on +5V reference voltage.
Hardly anything actually requires 24 bit A/Ds. Like weighing scales at supermarkets need accuracy like 10 kg, linearly.
However, most measurements really only want something like a 3 digit max plus a few decades of exponent. The 3 digits is for 0.1% of the reading (for resolution), and the decades switched at 1% accuracy.
I just roll my eyes at the original spec of 5% teamed up with a 24 bit A/D.
I have to say, that I am very surprized that Analog Devices, with their reputation, have produced such a dreadful BG reference. For less than 1 cent die area one can produce a BG (not zener) 2v5 reference with 20nV/sqrtHz with around 100 uA or so, although I do note that AD seem to require a few mA to achieve that performance in their products. I guess they are not hiring the right designers :-)
A buried zener can get better 1/f nose, but the applications I design for need to run on 2v7 not 7V! A good design can keep a BG to about a doubling of 1/f at 1 Hz from flatband.
24-bit A/Ds (actually 20-bit, but they spit out three 8-bit bytes) are chea p and widely available.
Why spend more on something that is harder to get hold of?
I was thinking that way back in 1993. It took me until 1996 to get the pape r written.
Sloman A.W., Buggs P., Molloy J., and Stewart D. ?A microcontroller
-based driver to stabilise the temperature of an optical stage to 1mK in th e range 4C to 38C, using a Peltier heat pump and a thermistor sensor? ? Measurement Science and Technology, 7 1653-64 (1996)
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510
http://electrooptical.net
http://hobbs-eo.com
Are the switched resistors in the same bridge arm as Rx ? Thus the switched resistors would be in the sub-milliohm range. Inserting a mechanical range switch with unpredictable resistance would destroy the accuracy and repeatability.
That's more complicated. A discrete A/D don't do all the work. The point is use a cheap meter to do all the processing and display. The signal just needs gained up reliably.
Yes, that sounds about right. Although, the linearity of the analog path shouldn't be much of an issue. Also, V(RREF+DUT) is not so easy to get, if both of these are 4-terminal devices - and they should be.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.