ohm meter rant

[johansen]

why the lack of high precision cheap ohm meters?

i have a 6000 count meter that isn't well known, but it uses the same dmm chip as several 60$ 6000 count meters that are well known.

and i find that the ohms portion is rather wacky.
the datasheet says the internal ic generates .63 volts and what i've been able to figure out (by having two meters)

is that the voltage generated (mine seems to output .643 volts) is connected in series with the 10M 1.11M, 101K, 10K, and 1K reference resistors. and also in series with a 900K resistor.
(figure 27, page 26)
http://www.ic-fortune.com/upload/Download/FS9922-DMM4-DS-11_EN.pdf

what this means is, when the meter cycles through the ohms ranges, internal mosfets ground those 5 reference resistors, and then the adc compares .63 volts with the volts across the standard resistance.. unfortunately this means that 0 ohms is .60 volts dumped on the standard resistor. so i think the meter can be "Zero'd" by adjusting the value of the 900K ohm resistor.. Something to look for on your prized 60$ meters...

anyhow, i know that this dmm chip has a 60mV voltage range, and they use it for the 0-6 amp range of the 60 amp shunt (derated to 10 amps, its a full 5 milliohms)..
so i'm a bit peeved that they don't use the 60mV range to deliver you a legit 0-60.00 ohm meter.

if the meter can handle it, you might be able to change that 900K ohm resistor to 90K and just remember to divide by 10
divide by 100 would be better.. it would be nice to have a 0.001 to 6 ohm meter.. in fact i think that's about the most useful range for largish rotating machines..

edit: i believe my understanding of the ohms section is not complete, as i find connecting the meter to a negative .745 volts yeilds 1.093K, 10K, 100K and 1M ohm on the 6K, 60K,600K,6M and 60M ohm range. it reads over on the 600 ohm range.

Edit2: changing 900K ohms to 9K ohms doesn't change anything except the diode range, it reads .67 volts for a 67mv drop.
i changed it back to 90K and the resistance range settles faster, diode range is still broken.

[neggles]

You might find that the meter is applying a constant current and measuring the voltage drop via the voltage divider resistors (to change range), rather than a constant voltage and measuring the current passed - that would explain this behavior, would it not?

[robrenz]

To measure accurately in the .001 to 6 ohm range you have to use 4 wire "Kelvin" measurements and you must use some means to cancel out all the thermal EMF's (thermocouple voltages at each dissimilar metal connection). The cheapest good accuracy low ohms can be had in a good hand held LCR meter that has 4 wire connections.  The AC excitation of the LCR automatically cancels the thermal EMF's. The only catch is high inductance resistances can give erroneous readings.

A short video example showing a LCR meter giving exact agreement to a calibrated high accuracy Micro Ohm meter.

[alanb]

why the lack of high precision cheap ohm meters?

Are there any situations where high precision and cheap go together?

[JoeO]

why the lack of high precision cheap ohm meters?

Are there any situations where high precision and cheap go together?

What is your definition of "high precision" and "cheap"?

But in general, no.  There is no free lunch.

[c4757p]

There is one - for a strict definition of precision (vs accuracy), frequency counters tend to have staggeringly huge precision. You'll still pay for accuracy.

[TMM]

You can do a 4 wire reading with two multimeters and a DC bench powersupply. If you set the supply to a few volts or so the thermal EMFs (millivolts) become insignificant - turn on the power just long enough for the meters to settle on a reading otherwise the resistor will heat up.

[robrenz]

You can do a 4 wire reading with two multimeters and a DC bench powersupply. If you set the supply to a few volts or so the thermal EMFs (millivolts) become insignificant - turn on the power just long enough for the meters to settle on a reading otherwise the resistor will heat up.

I don't think it is quite that simple, as an example 1V and .001Ohm needs 1000A drive current.  The average test current for micro ohm meters is 1A or 10A  so microvolt voltage drops are in play and still require thermal EMF's to be eliminated by offset compensated ohms or current reversal techniques.

Edit: You are correct that using a higher current will give larger voltage drops and the thermal EMFs will be a smaller percentage of the voltage. But averaging two readings where the current direction is reversed for the second reading eliminates the thermal EMFs so no special equipment needed. Also when you take the current accuracy and voltage accuracy of your two meters together you will find your calculated Ohms accuracy will be pretty poor.

[johansen]

http://zone.ni.com/reference/en-XX/help/375472A-01/switch/thermal_voltages/

I made a four wire ohm meter 0-.1 ohms, running about 22 milliamps on the nominal 3 volt battery using two 68 ohm resistors on either side of the opamp and adc.. such that full scale is about 2.4 millivolts across the resistance, which runs off the battery actually, so its just a ratio divider. the "meter" is theoretically 10,000 counts.. or .24uV per count.

so anyhow you can see the thermal emf from the connections if you warm them up with your fingers... but given the voltages and gold plated kelvin clips (which is what i have) .. its really not that bad, very easy to get 1% accuracy from .01 ohms to .1 ohms.. with .0001 ohm resolution. dipping down to 10% accuracy around .001 ohms. but that's worst case, i get a very reliable reading from one foot of 10 awg wire every time, provided i zero it on the same device which provided your resistor and kelvin clips are made from the same materials on both ends of it, nulifies the thermal emf error

this can be done for 7$ in parts.. at least two thirds of that is already present in all DMM's.

it would be very easy to put a tiny, el-cheapo constant current buck converter into a meter to get you 100mA which is going to cost you about half a volt in the diode drop, so that's 50mW, figure 50% efficiency in the buck converter so you're going to be pulling 30ma from a 3 volt battery.

100mA gets you something that is actually useable.

since most meters have additional connections for thermocouples and transistor hfe, you could build a four wire measurement without adding anything but a constant current generator, which is already present in the transistor hfe check, but its set for a completely worthless half a milliamp half the time.... just make the connections actually useable, and use the dmm leads to complete the four wire measurement

[AG6QR]

this can be done for 7$ in parts.

There's your answer.  That's over 10% of the retail cost of your meter, and if you add in the margin, it would increase more than that.

Most customers of a $60 meter don't even read the specs of what its accuracy is when measuring low resistances.  Also, many casual users aren't going to bother with four-wire techniques.  This wouldn't add value for customers who don't need it, so the meter's sales volume would drop, as most customers opted for the competitor's $60 meter that lacked this feature.

So it would move the meter into a special niche, for which the volume would be low.  That $60 price was only possible at high volume -- to make it a low volume niche product would increase base cost very substantially.

Maybe that moves the meter into the $100 price range, maybe more?  But as you point out, any user who understands the problem can do it himself for $7 in parts.  I've done something roughly similar for less than that, using a LM317 plus precision resistor to make a constant current regulator, while using a millivolt meter to measure the voltage developed by that current across a resistance.



As an aside, I believe most meters do use a constant current technique when measuring resistances.  The data sheet you linked to talked about how it detects continuity, but it didn't give details about resistance.  For continuity, it uses a constant .63V source, presumably to avoid turning on most semiconductor junctions.  But for resistance, I've hooked a meter up to a potentiometer, and then put another meter in series measuring current, and found that the current stays constant as the resistance changes, at least until the autoranging decides to switch ranges.  I don't claim all meters do it that way, of course.

Regardless of whether the meter holds current or voltage constant, the limiting problem for high precision measurement of low resistance in most meters is the fact that they're using two wires, and not a four wire kelvin technique.

[sync]

Does your circuit withstand 1000V?
If not redesign it to make it feasible for a DMM.

[johansen]

Does your circuit withstand 1000V?
If not redesign it to make it feasible for a DMM.

my low ohms meter is one of those 7$ 1KG drug scales.. like i said, 7$ in price parts but two thirds of that is already in every dmm.
you're paying for kelvin clips essentially.
to make it withstand 1000 volts across the sense leads would require a handful of smd resistors and clamp diodes.. like every other meter does it.

the 6000 count meters in question already have 10 microvolts per LSB but its not available except in the 0-6 amp range through the .005ohm shunt.
They have a 60mv range as well but i haven't seen any dmm that has that available.
fluke does, of course, and some of their meters have low ohms that goes down to .01 ohm.. so for their 50,000 count meter that's 500.00 ohms

how hard can it be to push the current source up one decimal point?

in any case, all you need is a 100ma ccs and you've got a 0-.6 ohm meter in .001 resolution with these cheap dmm chips.

if you want .0001 ohm resolution you need a chopper stabilized amp or a very low offset opamp that can be trusted. i'm not sure which is cheaper to implement these days, and of course you need kelvin clips.

with 14 gauge test leads (now there's a good idea!) you don't really need 4 leads to measure 1 milliohm.
its not like you have 1milliohm of variable contact resistance for clean nickel plated test leads, in practice its one tenth of that, just stab the points in harder...

[robrenz]

IMO If you want reasonable accuracy below 1 ohm you really need to use 4 wires and active offset compensation or current reversal. 14 ga leads do not solve all the problems. 

[sync]

the 6000 count meters in question already have 10 micrvolts per LSB but its not available except in the 0-6 amp range through the .005ohm shunt.
they have a 60mv range as well but i haven't seen any dmm that uses it.

UT61D for example.

in any case, all you need is a 100ma ccs and you've got a 0-.6 ohm meter in .001 resolution.

Show me a 100mA CCS suitable for a hand held DMM which withstand 1000V at the test leads. The DMM must not explode or catch fire. It's not easy and probably cost more than 7$.

In the FS9922 data sheet they use a 1.2k PTC resistor for protection. At 100mA it will have a voltage drop of 120V! And then heat up and increase it resistance.

[alm]

with 14 gauge test leads (now there's a good idea!) you don't really need 4 leads to measure 1 milliohm.
its not like you have 1milliohm of variable contact resistance for clean nickel plated test leads, in practice its one tenth of that, just stab the points in harder...

This highly depends on the component. Leads may be corroded. You can't very well stab the points into 1/4 W resistor leads, or SMT resistors. Contacts often make decent force transducers. The resistance of pressing my Fluke leads together varies from something like 17 mOhm when I apply little pressure to down to ~10 mOhm when applying a fairly large force. Do you want your measurements to depend on the amount of force you apply?

I would definitely use Kelvin connections for measurements below 100 mOhm, or if I need better than 10 mOhm resolution.

[johansen]

Show me a 100mA CCS suitable for a hand held DMM which withstand 1000V at the test leads. The DMM must not explode or catch fire. It's not easy and probably cost more than 7$.

I don't believe any meter i've seen will pass that test for 1000 volts applied to either the thermocouple port, or the transistor hfe port.
also, does that test apply to the 10 amp range which typically has a sand filled ceramic fuse?
by that standard, all you need is two 1 amp fuses to protect your 100ma ccs from all possible failure modes.
if you really want to get insane, you could isolate the 100ma ccs through a tiny transformer, and just full wave rectify, and then you only need one fuse.

[alm]

I don't believe any meter i've seen will pass that test for 1000 volts applied to either the thermocouple port, or the transistor hfe port.
also, does that test apply to the 10 amp range which typically has a sand filled ceramic fuse?

Have you seen any decent meters? No good meter has a socket for transistor testing, and thermocouples are generally connected to the voltage terminals. The 11A fuse in Fluke meters is rated for 1000 V and can interrupt currents up to 20 kA. IEC61010 requires meters to be safe at their rated voltage at any input and setting.

by that standard, all you need is two 1 amp fuses to protect your 100ma ccs from all possible failure modes.

That might work with anti-parallel diodes diodes across the current source. You've just added in excess of $7 in protection circuitry to the BOM costs, though, and a fair amount of PCB real-estate.

if you really want to get insane, you could isolate the 100ma ccs through a tiny transformer, and just full wave rectify, and then you only need one fuse.

Are you measuring current with an AC signal, or are you suggesting an isolated DC-DC converter?

[johansen]

IEC61010 requires meters to be safe at their rated voltage at any input and setting.

Good thing i can't afford that specification then    regulations should be free if they have any legal authority.

Anyhow, I suggested an isolated dc-dc converter. you can set and forget the input side of the converter to set the current.. calibrating itself is easy enough, just clip the leads back into the amp meter socket.. 1Kvac isolation is trivial.. though you need 3mm iirc for creepage.

[Wytnucls]

There are cheap milliohm meters available. Here is one made by CEM, equipped with Kelvin clips:
Best accuracy 1mOhm.

http://www.ebay.com/itm/CEM-DT-5302-High-Accuracy-Kelvin-4-Wires-Milliohm-Meter-/220839780573

[robrenz]

That meter has a max 100% of reading error at .001 Ohm  A .001 Ohm resistance could read zero or .002 Ohm and be in spec.
That does not mean it is junk, it just points out that it is difficult to find meters that are high accuracy as in percent of reading errors when you get below 1 ohm or even worse below .001 Ohm. The Cambridge tech/IET 510A Micro-Ohmmeter is the best I have found for a reasonable price. If you are patient/bold enough they can be had for under $300.00 for a meter that is over $2000.00 new.  This post from my thread on the LOM 510A shows how you must do a percent of reading spread sheet to make any sense of comparing specs. You can see that different meters can do well in certain ranges and then be much worse in other ranges. The 3458A shows this in being the most accurate until 0.16 ohm where the 34420A and the LOM 510A are much more accurate. FWIW

[mzzj]

There is one - for a strict definition of precision (vs accuracy), frequency counters tend to have staggeringly huge precision. You'll still pay for accuracy.

Anything to do with frequency is relatively dirt cheap.

You can get 1ppm accurate oscillator for couple of bucks. Now try that with pretty much anything else like voltage, current, lenght or god forbid 1kg calibration weight 

[Neilm]

Good thing i can't afford that specification then    regulations should be free if they have any legal authority.

Paid for by whom? They are international standards. If you are a company, £70 for a standard (less if a member) is a very minor overhead. It goes on helping publish the standard. The new version of 61010 was discussed for about 5 years prior to publishing. The work for that is paid for by the professional organisations of the countries in that committee.

[Matje]

Good thing i can't afford that specification then    regulations should be free if they have any legal authority.

Paid for by whom? They are international standards. If you are a company, £70 for a standard (less if a member) is a very minor overhead. It goes on helping publish the standard. The new version of 61010 was discussed for about 5 years prior to publishing. The work for that is paid for by the professional organisations of the countries in that committee.

Well, I'm sure something could be arranged, at least with the standard organizations where the members are countries, or rather the national standards orgs. BTW, it's not always that cheap, I have an oldish SQL standard around at work that cost a couple of hundred (700 I seem to remember), hmm, D-Marks I think it was back then.

Don't be fooled, some of the companies printing/selling standards make a metric boatload of money off it, as they often have a national monopoly.

In my opinion this goes directly against the intention of standardizing stuff: If you minimize the number of people/companies/entities able to actually read and thus adhere to a standard - general fail ensues. Of course that is the way the big players like it...

[Wytnucls]

That meter has a max 100% of reading error at .001 Ohm  A .001 Ohm resistance could read zero or .002 Ohm and be in spec.
That does not mean it is junk, it just points out that it is difficult to find meters that are high accuracy as in percent of reading errors when you get below 1 ohm or even worse below .001 Ohm. The Cambridge tech/IET 510A Micro-Ohmmeter is the best I have found for a reasonable price. If you are patient/bold enough they can be had for under $300.00 for a meter that is over $2000.00 new.  This post from my thread on the LOM 510A shows how you must do a percent of reading spread sheet to make any sense of comparing specs. You can see that different meters can do well in certain ranges and then be much worse in other ranges. The 3458A shows this in being the most accurate until 0.16 ohm where the 34420A and the LOM 510A are much more accurate. FWIW

I get the impression the OP was after a portable multimeter with an accurate reading on low Ohm ranges, while at the same time providing the usual voltage, current and capacitance measurements. Of course, dedicated milliohm bench meters connected to the mains, will be far more precise, but with a hefty price hike and bulk. This one can only produce a maximum of 200mA from its 6 AAA batteries.

[sync]

It costs about 130$. Not 60$ + 7$ in parts as the OP wanted. Has only a basic accuracy of 1% and is probably unsafe. That are not 1000V fuses.