How to measure low resistance reasonably accurately?

t

ted

upid

rom pure DC measurements.

op amps like ADA4522 or OPA189, with good thin-film or bulk metal foil resi stors. As for Sigma-Delta ADCs - the champions are currently in the AD7xxx family - AD7177-2 or the newer AD7768. To get more than 20 meaningful bits you need an external buried zener reference. Or if you go for ratiometric s etup - these ADCs can output their internal Vref. They come with tons of ot her features too.

ished by the same guys:

Here's a better link:

formatting link
esistance_Comparator

Cheers, Nikolai

Reply to
Castorp
Loading thread data ...

Okay, they do the averaging for you. Potentially quite useful. The analogue LPF is going to be necessary for EMI resistance and to prevent RF junk from aliasing down into the fundamental interval.

Sine wave AC techniques get a bit more difficult down in the very low ohms region on account of inductance. An inch of wire (~20 nH) has 1.25 milliohms of reactance at 10 kHz, and its RL time constant isn't that quick--remember that the time constant is L/R, so that inch of #10 AWG wire (0.003 ohms/m or 76 uohm per inch) has a time constant of

TC = L/R = 20 nH / 76 uohm = 260 us.

Slow square wave or switched-DC eliminates the problem, because the RL time constant is still pretty short--it'll settle out in a few milliseconds. JL talks about eddy current problems, which I imagine could be longer-lived than that because the resistance is much lower.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

d

id

Bob Parker's Blue/EA ESR meter kit reads to 0.01R. Perhaps not enough in

this application but I'm putting it out there in case someone is interested...

John :-#)#

--
(Please post followups or tech inquiries to the USENET newsgroup) 
                      John's Jukes Ltd. 
 Click to see the full signature
Reply to
John Robertson

At such high frequencies you should also consider the skin depth, which will have some effects on conductor diameters larger than 2 mm.

Reply to
upsidedown

I've seen voltage overshoots from a current step caused by eddy currents. Numbers like tenths of a per cent with taus in the hundreds of microseconds. We invented an apparently eddy-free current shunt, a shaped strip of annealed manganin bonded to an aluminum heat sink.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

** Yep - Bob's ESR & Low Ohms Meter is my "go to" in the first instance - mostly to make sure I am not misreading the colour code on the resistor.

However the test frequency used is 100kHz, enough to invoke significant inductance effects due to the turns on a WW type or the spiral cut on a film type.

.... Phil

Reply to
Phil Allison

Axial lead thin film resistors are spiral trimmed. Surface mount planar thin film resistors are trimmed with an "L" cut - the initial cut across the current flow does the coarse trim and the second cut, parallel to the current flow, does the fine trim.

You get lower inductance and lower parallel capacitance that way - which gave us a bit of shock when we first ran into it, back around 1985.

--
Bill Sloman, Sydney
Reply to
bill.sloman

I'm just tossing out an alternate approach to the others which seem to be mostly about finding the right awesome-sauce hardware; if one's going to build a design that uses a general purpose computer anyway I'm a fan of leveraging processing power to compensate for consumer-grade hardware, not buying awesome-sauce hardware to provide the computer with excellent data and make its life easy.

Something like a hidden Markov monte carlo technique to find the value of the hidden variable (the true resistance) from observables corrupted by things like variations from Joule heating, ADC nonlinearity errors, etc.

It is probably overkill for a workshop tool but something to think about

Reply to
bitrex

Repeat after me: "Postprocessing is not a substitute for good data." (Except in science that fits someone's narrative, of course.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

I already own 2 meters able to go down to 10mOhms, but for me it is the very beginning where the interesting stuff happens. With a bench PSU set for 1A in CC mode I can go down to 1mOhm (reliably), would like to improve it by a factor of 10 and automate all the offset-related problems. Doing it by hand is tedious.

Best regards, Piotr

Reply to
Piotr Wyderski

I like the idea, thanks, Nikolai! Not sure why they went to the very end of the analogue route, the final S/H circuits could have been eliminated and replaced with a properly written piece of code.

Best regards, Piotr

Reply to
Piotr Wyderski

IMHO the number of useful bits has nothing to do with the accurcy of the Vref source. If it is accurate down to, say, 0.1%, the bits might still be useful (that is, stay monotonic or even linear) for all the processing purposes, just the final value will be in the relative accuracy 0.1% ballpark, too. In other words, Vref's source noise level is the true limiter, not its accuracy. One can have a very useful tool with robust nanoohm-range resolution, just not nanoohm accuracy.

Yes, the paper is great!

Best regards, Piotr

Reply to
Piotr Wyderski

The Serbian paper shows how to get rid of offsets with an unipolar current source.

OK, I think my starting point should be a 4-terminal

200mOhm 0.1% resistor (~10USD from Mouser) in series with a 4.7Ohm resistor (to limit the short-circuit current down to ~1A assuming a 5V PSU and move this heat producing part away from the 200mOhm reference) then in series with the DUT and with a MOSFET to gate the current. That way I will be able to measure U_DUT|I=max, (U_DUT+U_RREF)|I=max, U_DUT|I=0, (U_DUT+U_RREF)|I=0 and do the offset compensation the Serbian way.

Best regards, Piotr

Reply to
Piotr Wyderski

If you just want the low resistance feature that cheap meters don't have, build an amp/sensor system and feed the output to one of those cheap meters. Have a look at chopper op-amps. Probably some differential ones about...

Waste of time reinventing the wheel to use another A/D.

-- Kevin Aylward

formatting link
- SuperSpice
formatting link

Reply to
Kevin Aylward

With limited currents (100 mA was mentioned) the autozero op amps don't help: thermocouple effects just dominate. Doing current reversal DOES help, and even non-autozero op amps are accurate for the alternating voltages that result.

A good scheme might be diff amps on both a reference resistor, and the unknown resistor, carrying same current in series; then use three or four cascaded low-gain stages, and do your A/D conversion on the unknown's highest gain output that isn't saturated.

I'm assuming 1 ohm down to 1 milliohm, and 1/10 amp drive. You can go lower with the AC scheme than that.

Reply to
whit3rd

The audio-engineering equivalent saying is you can't fix a bad take "in the mix" by e.g. using EQ and dynamic range compression and pitch correction, it'll just be a bum vocal take. that's been EQed and compressed and pitch-corrected.

Don't know that this is a comparable situation there's good data there it's just been masked. There's no masked "good take" in a wildly off-key vocal recording. Or even some type of extrapolation or average of a thousand of them.

The Space Shuttle didn't fly on great IMU data by modern standards but fly it did. Or at least the times it unfortunately didn't weren't at all related to uncertainty about its position.

Reply to
bitrex

The issue is whether they stay _still_. Most delta sigmas will only stay still to 18-19 bits even with their inputs shorted. Some will do better if you run them very slow--10 Hz updates or something like that. The rest of the bits are decorations courtesy of the marketing department.

In a true ratiometric measurement, the reference's noise isn't a problem either. If the frequency response of the reference and signal paths are different (which they generally are), the deviation from ratiometric operation will cause noise problems.

I'll have a squint at it.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Nope. That just means that the track or data isn't what you hoped for. Deferring to personal preference is perfectly okay in audio, but not in science.

There are times when the very best data you can get still aren't as good as you'd like, so you have to pile on the postprocessing to try to get something usable.

Ohmmeters aren't in that class, even microohmmeters.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

The Tongui TH2822A/C/D/E series LCR meter goes to 100 uOhm as well as providing inductance and capacitance measurements accurate to 0.25%.It uses Kelvin clips that are different than the ones used by the TH2821 series. So if you bought the TH2821 meter and want to upgrade, you need to get a new set of Kelvin clips. But the upgrade may be worth it. The TH2822 series has many more features, including averaging. This will help greatly when measuring close to the limits of the specs. See

formatting link

Available on eBay at US $360.46:

262997838423
Reply to
Steve Wilson

If your measurement outputs meet whatever your standards for accuracy and acquisition speed are and you arrive at those outputs via some fashion of post-processing rather than higher linearity ADC, or better-specced op-amps, say, and the overall design and implementation costs end up being less doing it the former way rather than the latter then objections as to the particular nature of your input data as being "good" or "bad" would seem to be all existential ones

Reply to
bitrex

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.