How/when/why did meters get that 1/2 digit?

[brucehoult]

As in the ubiquitous 3 1/2 digit DMM.

Of course it's not really half a digit. It's 1 bit, or 0.3010 of a decimal digit.

It's not that useful for measuring random things, as on average only ... 30% ... of values lie between 1 and 1.999.

Maybe designed things are more likely to have values near to 1, 10, 100. Certainly 12V or 110V are common in certain applications. But no advantage at all if you're looking for 6V, 24V, 48V, 240V. Fortuitous maybe that the size of the Volt was picked to be in the range of the potential generated by a single cell for many chemistries -- or near inevitable?

I guess the biggest thing that can be said about a "1/2 digit" is that's it's cheap to add. Just a single LED/LCD segment (though you might make it look like two, or it might even be cheaper to build it as two).

[pelule]

I recommend to use the search function here in eevblog.
This item was discussed already several times and in detail.
/PuLuLe

[Nusa]

It's not all about the display. It's also about being able to display close to the precision the meter is capable of.

https://bkprecision.desk.com/customer/portal/articles/1535307-difference-between-counts-and-digits

[CatalinaWOW]

This may be a case of convergence.  Two different reasons giving the same answer.  As I recall from the trade journals when digital displays were first becoming a thing the seven segment LED display and associated driver was a significant cost element.  Thus a simpler left digit.  No decoding, just a bit on or off.  Only two segments and possibly a decimal point that have to work in the display.  (Allowed productive use of quite a few rejects).  Add that to the better match between measurement accuracy (counts) and display capability and you have a winner.  As time went on LEDs got cheap, drivers got integrated and LCDs even more integrated took over.  The cost reasons virtually disappeared but the marketing strategy was already in place.  Just like counting the number of tubes (valves) in a radio or TV carried over into counting transistors as a figure of merit.  Until integration finally made that even sillier than it already was.

[Gyro]

The last discussion on the subject...

https://www.eevblog.com/forum/testgear/relationship-between-dmm-resolution-and-'counts'/

[PA0PBZ]

Having solved that can we now talk about 2.5 channel scopes? 

[pelule]

I still have an issue with a 1/2 wire. 

[martinr33]

In accounting, numbers are far more likely to begin with a 1 than any other number. This is known as Benford's law. A quick extract from Wikipedia:

It has been shown that this result applies to a wide variety of data sets, including electricity bills, street addresses, stock prices, house prices, population numbers, death rates, lengths of rivers, physical and mathematical constants,[3] and processes described by power laws (which are very common in nature). It tends to be most accurate when values are distributed across multiple orders of magnitude.

Seems like this is a good description of what we see. It also makes some sense of the 2V prefix on some Keithley meters.

[brucehoult]

In accounting, numbers are far more likely to begin with a 1 than any other number. This is known as Benford's law. A quick extract from Wikipedia:

This is not news to me, and was explicit in my "Of course it's not really half a digit. It's 1 bit, or 0.3010 of a decimal digit" in the original post.

To use slightly different words, it is not that half of all numbers start with a 1, only about 30% (0.3010) do. This is because log(2) - log(1) = 0.3010. Similarly, log(3) - log(2) = 0.1761 or 17.6%; log(4) - log(3) = 0.1249 or 12.5% etc.

You see these figures illustrated in that wikipedia article on Benford's law, though without mathematical explanation.

[brucehoult]

I recommend to use the search function here in eevblog.
This item was discussed already several times and in detail.
/PuLuLe

I did, thank you. None answered my question.

You can make a DMM with 1000 counts. I don't ever recall seeing one. Maybe they happened for a brief period.
You can make a DMM with 2000 counts. Maybe 90%+ of the meters on the market are like this.
You can make a DMM with 4000 counts. Some exist, but it seems to be quite rare.
You can make a DMM with 10000 or 20000 counts. That exists but is very high end.

It should be pretty much exactly as much harder to make  4000 count DMM compared to 2000 count, as to make a 2000 count DMM compared to 1000. But *everyone* takes that first step -- you just don't see 3 digit DMMs -- and almost no one takes that 2nd step.

That seems like it should have a psychological/marketing explanation rather than a physical/electronic reason.

Sure, a 2000 count DMM is better than a 1000 count one, but a 4000 count DMM is exactly as much better again ... another 30% of numbers start with a 2 or a 3.

Was there ever a time that 3 digit DMMs were in the majority? I don't remember it. Did DMMs start from 3 1/2 digits, right off the bat? Or was someone first and then shortly after you simply couldn't sell a 3 digit DMM at any price? And yet there's been very little pressure for the low end to move from 2000 to 4000 count. It's weird. It's *got* to be psychological. not cost or difficulty.

[PA0PBZ]

Don't forget the massive popularity of the ICL7106/7107 back then, and I guess the 2000 count was a compromise between price/performance/complexity.

[Kleinstein]

There is a little second effect that make DMMs with something like 4000 counts less attractive, at least for the simple circuits. For the current measuring shunt a voltage drop of about 200 mV is still acceptable, but a 400 mV drop can be a problem. So simple 2000 count meters have a 200 mV / 2 V / ... range for voltage and measure current with shunts of 0.1 / 1 / 10 ...  Ohms so they can use the internal 200 mV range. With a 400 count meter there usually is 400 mV as the lowest range, but a 400 mV drop on the shunts is more on the low quality side. So a good 4000 could meter would need an extra amplification stage to get a 200 mV (or lower) burden.

One point to having so many 3,5 digit meters is also that there is are rather successful ICL7106/7 chips, that was used in a large number of early DMMs. So there was little incentive to settle for less and it made lower resolution chips obsolete quite fast.

With the higher end meters there is the slight tendency to have ranges to around 12-15 V, since normal OPs / CMOS switches can work reasonably good up to about 15, maybe 18 V.

[Wytnucls]

You can make a DMM with 1000 counts. I don't ever recall seeing one. Maybe they happened for a brief period.

1975 1100 counts HP3476:

[Nusa]

I recommend to use the search function here in eevblog.
This item was discussed already several times and in detail.
/PuLuLe

I did, thank you. None answered my question.

You can make a DMM with 1000 counts. I don't ever recall seeing one. Maybe they happened for a brief period.
You can make a DMM with 2000 counts. Maybe 90%+ of the meters on the market are like this.
You can make a DMM with 4000 counts. Some exist, but it seems to be quite rare.
You can make a DMM with 10000 or 20000 counts. That exists but is very high end.

It should be pretty much exactly as much harder to make  4000 count DMM compared to 2000 count, as to make a 2000 count DMM compared to 1000. But *everyone* takes that first step -- you just don't see 3 digit DMMs -- and almost no one takes that 2nd step.

That seems like it should have a psychological/marketing explanation rather than a physical/electronic reason.

Sure, a 2000 count DMM is better than a 1000 count one, but a 4000 count DMM is exactly as much better again ... another 30% of numbers start with a 2 or a 3.

Was there ever a time that 3 digit DMMs were in the majority? I don't remember it. Did DMMs start from 3 1/2 digits, right off the bat? Or was someone first and then shortly after you simply couldn't sell a 3 digit DMM at any price? And yet there's been very little pressure for the low end to move from 2000 to 4000 count. It's weird. It's *got* to be psychological. not cost or difficulty.

The first really successful digital multimeter was probably the 3.5 digit Fluke 8020a starting in 1977. Anything that existed prior to that probably wasn't robust or good enough to replace the good analog meters of the day. So no, any 3 digit meters that existed never made it big.

I disagree with your assessment of "very high end", which I'd call 50,000 counts and up. Even higher for benchtop versions. See eevblogs excellent coverage of professional meters on the market: https://www.eevblog.com/forum/testgear/multimeter-spreadsheet/

If you were to make a DIY multimeter based on an Arduino analog input, it would have 1000 count accuracy. It's a pretty easy project, circuitwise.

[Andreas]

Hello,

and in the end it was probably only a "marketing gag"
RCA had a chipset CA3161/CA3162 capable of displaying 3 digits: -99 to 999 counts (total 1099 counts including zero).

So perhaps Intersil with the ICL7106/7107 only wanted to have double resolution.

With best regards

Andreas

[try]

Andreas,

and in the end it was probably only a "marketing gag"
RCA had a chipset CA3161/CA3162 capable of displaying 3 digits: -99 to 999 counts (total 1099 counts including zero).

So perhaps Intersil with the ICL7106/7107 only wanted to have double resolution.

is that Intersil or rather you starting to sum up negative figures as well?
I am somewhat confused.
Negative figures have never been included into the count calculation, as far as I am aware of.

Regards
try

P.S.: corrected typo error

[Andreas]

I only wanted to mention that the negative count was much lesser than the positive at that time.

Andreas

[Cerebus]

I disagree with your assessment of "very high end", which I'd call 50,000 counts and up.

The volt-nuts over in Metrology call that 'entry level'. My latest acquisition, a 220,000 count used Keithley 197 turns up later this week.

[bitseeker]

I disagree with your assessment of "very high end", which I'd call 50,000 counts and up.

The volt-nuts over in Metrology call that 'entry level'. My latest acquisition, a 220,000 count used Keithley 197 turns up later this week.

Depends if you're talking about handheld or bench DMMs. 50,000 to 500,000 counts is up there for handheld meters. For bench ones, 1M counts is readily available and is probably entry level voltnuttery. 10M starts to get pricey and is more like a good start into voltnuttery. 100M counts is high end and there aren't many to choose from.