What are 6.5 Digit Multimeters Good For?

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

Calibration source. You want at least 4x better accuracy than the DUT.

but he was referring to millohm measurements using a Kelvin connection.

--
Jim Yanik
jyanik
at
localnet
dot com

I think people with three place data and churning it to "accuracies" of 4 or 6 places are nuts equivalent to climate change scientists :-) ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
      The only thing bipartisan in this country is hypocrisy

We got calculators and computers with big registers and mantissas these days. Keep all the accuracy possible along the entire observation chain, and do the rounding off at the end.

They last about 25-30 years with the odd bit of refurbishment.

A 7 DVM model still in use today (probably one of the last manufactured that way) is in the images online at Southampton university.

Regards, Martin Brown

This is a good strategy if you don't want to be bothered with keeping track of the number of significant figures in your results along the way.

The main downside is that it can lead people to just blindly trust their calculators when they need an answer accurate only to, e.g., 3 or 4 digits, but don't realize that in their calculations somewhere they performed an operation that required more like 15-20 significant figures, and therefore their answer is completely inaccurate. The most common example is perhaps "small differences of large numbers" -- one minus (one plus 1e-20) is just zero on most calculators. The "canoncial" expression for one root of a quadratic has this problem: -b + sqrt(b^2-4ac) ...

Of course, there are ways to fix all this, my point is just that one shouldn't implicitly "trust" a calculator just because it has far more digits than you actually need in your answer.

---Joel

It all depends on what you are doing. Opinions vary because needs vary. 95% of the time my needs are filled by the small stack of Fluke 4 1/2 digit handheld DMMs I have. They are high quality, accurate, reliable, robust, easy to use, electrically floating, and the lack of 'weirdness' means I trust them a lot. They are built like tanks and you can get good used ones on ebay for $50-$100.

I rarely use the '5 euro' type meters unless I need something expendable. Although cheap meters work fine for the vast majority of uses they can sometimes lead you astray with accuracy drift, injecting noise or current into sensitive circuits, or just erroneous measurements because they don't react predictably to weird waveforms. If they are poorly shielded they can pick up RF and inductive fields and give you strange results too. I wasted days troubleshooting once in China because the accuracy and linearity of my cheap meter was thrown off by the high humidity. There is a lot of careful engineering that goes into making an accurate high quality instrument that stays accurate and that reacts predictably with all signals and isn't affected by weird environments. I have enough problems without chasing down problems coming from my test gear. High quality instruments can save a lot of time.

For that other 5% of the time I have an Agilent 34410A 6 1/2 digit DMM. I use it when I need to monitor millivolt or smaller changes on nodes that are 5-10 volts away from ground. When you are working with sensors signals can be small, gains high, impedances high, currents low etc. and it can all be riding on relatively high voltage levels.

Many times it's not so much the absolute accuracy or the 6 1/2 digits but the usable measurement ranges are extended. Being able to measure up to 1 Gohm can be helpful in tracking down board leakage problems, humidity or contamination problems, or just tracking down problems with high impedance circuits. Resolution below the milliohm level and the 4 wire low ohms measurement ability is handy for tracking down PCB problems and high current circuits. At the other end of the scale sometimes you need to see what's going on down at the uVolt scale where thermal coefficients and thermocouple effects can be an issue. You can identify and understand thermal effects without needing large temperature swings. Sometimes you want to see what's happening in the nanoamp range. This can be useful for tracking down current leakage problems, or photo-current problems from light getting to your silicon from bad packaging (like when your product works great in the lab but fritzs out every time you try to demonstrate it in the brightly lit conference room with the big picture window and the scowling executives). Sometimes a circuit is running around the 1 micro amp level and you need enough resolution to see what is happening when conditions change. Often it's not absolute measurements that you need, it's just the ability to 'see' what's going on at small scales, or to see small trends.

High input impedance in the 10V scales and under is another very powerful feature that has nothing to do with it being a 6 1/2 digit meter but comes with that class of instrument. Standard 10M-20M ohm input meters are troublesome with high impedance circuits. They pull bias points around, inject current and noise, probe capacitance can make things oscillate and change frequency responses etc. With an input impedance > 1Gohm you can put a 100k-1M resistor in series with your probe to isolate it and still get accurate measurements without affecting your circuit.

These are just some of the things I find them useful for. Other people get a lot of value out of the programmable measurement stuff for automated and long term measurement needs. These instruments are a whole collection of useful features besides the accuracy and resolution. Sure you can do some of these things with voltage and current sources and other work arounds but it's awkward and they introduce problems of their own.

As I said at the beginning, it all depends on what you doing. If you can't think of why you need one then you probably don't. When you do need a 6 1/2 digit meter, you'll know it. Going from a 3 1/2 digit to a 4 1/2 digit will give you some idea of the advantages of higher resolution. For me I did fine with 4 1/2 digits for years but I eventually got into stuff where 6

1/2 was a worthwhile investment. I don't ever expect to want an 8 1/2 digit.

My recommendation is don't buy a 6 1/2 digit meter till you find yourself in a situation wishing you had one to make your life easier. Like true love you'll just know it when it happens, but that time may never come.

Measure the sheet resistivity of traces on a few of your boards and tell us what you see.

John

You do know that we do not use copper boards any more, right?

I didn't know that. What do you use?

John

It is an RoHS world, John. What do you think we use?

It ain't HASL over Copper or SMOBC, I'll tell ya.

Can't imagine.

You said that "we do not use copper boards any more." So what are the traces made out of?

John

What does RoHS have to do with Copper, AlwaysWrong?

IOW, you don't know.

The only place copper is plated is in the holes. The 1 oz copper layer for traces is a copper sheet that is laminated to the board. It's rolled out to the proper 1.4 mil thickness when it's manufactured. It's not plated on by the PCB manufacturer.

You may be making measurement errors when you measure the traces, 1 amp can be a lot of current on small traces and you may be heating that trace up quite a bit. Copper has a substantial thermal coefficient of resistance, heat it up by 25C and the resistance goes up 10%. Try dropping your current to 0.1A and see how your measurements change.

--------------------------------------- Posted through

Oh no! Don't ruin his bent belief that he has been getting ripped off all these years!

His bent perceptions of reality must be kept in place or he'll explode, taking half the city with him!

Fields

BC

:

snipped-for-privacy@yahoo.com=20

Fields

from BC

:

really=20

chip,

around

cheap

accurate=20

antiqued=20

measured.

in

As far as i know they were relegated to primary standards labs=20 some 30+ years ago. I would bet on not finding them in anyplace=20 else (any more), maybe only at NIST class calibration reference=20 services now.

snipped-for-privacy@yahoo.com=20

BC

:

chip,

++++expensive stuff anyways.

or ITER, or LIGO,

Perhaps i wanted to get credible results on contact resistance=20 repeatability. Or maybe i need to verify a current measuring shunt.

I think my measurements are accurate, and I often include a test trace whose geometry is suitable for accurate sheet resistance measurement.

If I include a fab note demanding a minimum sheet resistance, or say "START WITH 1 OZ COPPER" I usually get below 600 uohms/square. If I just say "COPPERCLAD 1 OZ" I usually don't.

John

can

current

Well if you are sure you are not heating your traces up when you measure them it sounds like you need to talk to your fab about the issue. Get them to explain their fabrication process and what the variables are that can affect the sheet resistance. 1 oz copper is just an arbitrary standard, what's important is to know what sheet resistance you can expect out of them and how variable it can be. It's good to know what the sheet resistances and dielectric constants of all the board options they offer so you can get what you need out of them. If it's an important aspect of your board design you have to specify it, you can't leave it to chance since they may change suppliers, or processing, or subcontract it out etc. Talk to them, find out what the options are and what they are comfortable with. If you are not satisfied find another fab.

What fab are you using?

--------------------------------------- Posted through

We use several board houses. I am used to the concept that a "1 Oz" board may have 750 uohms per square, and on most boards it doesn't matter. If it does matter, as when high currents are involved, I make sure the fab notes clearly demand true 1 Oz, and include a test trace.

When I do include such a note, I usually get a phone call about it.

John