We Can Make Multimeter Reviews Better

[EEVblog]

....
This is why to me, arguing over whether or not a meter survives Joe Smith's overload tests slighter better than some other meter is likely moot for most people. ....

Sure, drag me into this mess...
I don't check meters for their accuracy as part of my reviews.  I don't often check all their features.  Rare I even look at current measurements.  Rather I focus on stressing the meters to failure.  Life cycling their rotary switches, exposing them to chemicals, dropping them, and of course subjecting them to some low level transients.    It all takes time.   
I don't normally accept donations so there is the cost of the product (which is going to be trashed).  But that pails to the time required.   No Pateron, begging people to join the channel,  ring that bell or what ever YT wants.  It's a total loss channel.   Then the content is sparse and dry.   
I run the tests I want to see and make them available for free.  That about sums it up.  Don't like the content, plenty of it out there.

To be clear, I like your content Joe, I find it interesting and valuable. Just pointing out that probably for most people, something like high voltage overload testing comparisons are a bit moot for most once you pass a certain generally acceptable point. So it's an example just like the OP is discussing here with low signal level AC. Sure, small signal AC would be an interesting test and comparison, but for most people it's probably moot in the decision making process.

[joeqsmith]

....
This is why to me, arguing over whether or not a meter survives Joe Smith's overload tests slighter better than some other meter is likely moot for most people. ....

Sure, drag me into this mess...
I don't check meters for their accuracy as part of my reviews.  I don't often check all their features.  Rare I even look at current measurements.  Rather I focus on stressing the meters to failure.  Life cycling their rotary switches, exposing them to chemicals, dropping them, and of course subjecting them to some low level transients.    It all takes time.   
I don't normally accept donations so there is the cost of the product (which is going to be trashed).  But that pails to the time required.   No Pateron, begging people to join the channel,  ring that bell or what ever YT wants.  It's a total loss channel.   Then the content is sparse and dry.   
I run the tests I want to see and make them available for free.  That about sums it up.  Don't like the content, plenty of it out there.

To be clear, I like your content Joe, I find it interesting and valuable. Just pointing out that probably for most people, something like high voltage overload testing comparisons are a bit moot for most once you pass a certain generally acceptable point. So it's an example just like the OP is discussing here with low signal level AC. Sure, small signal AC would be an interesting test and comparison, but for most people it's probably moot in the decision making process.

I wonder when we talk about high voltage overload, what you personally consider an acceptable level to be?   I've talked about how dry our house gets in the winter months and ESD events become common place.   The simple act of walking to the chair and sitting down can create several kV transients.   At least for me, I wouldn't consider grounding myself before using a handheld DMM. 

The low energy wide pulse testing I perform are a bit of a guess.  Seeing that $50 Amprobe AM510  survive 5kV pretty set the criteria for what I consider an acceptable level.  Any meter that would survive the tests on that digital transient generator, I consider robust.   If it falls short, or worse, can't be repaired, I consider it a poor design.   

If a handheld meter breaks down internally on that same generator because of a poor layout, bad part selection,  bad mechanical design, it's nothing I would ever consider using on a high energy circuit.  The UT90A is a great example.  That design is so bad, that meter is still around today.  It will breakdown and absorb all the energy I can throw at it and save the delicate control ICs...   See time stamped:

https://youtu.be/aRuI_q_K5RY?t=409

[joeqsmith]

...
Accuracy
Its not uncommon to find that a very expensive high-precision bench multimeter reading dc to better than 0.03% has an accuracy of worse than 1% for ac voltages. A much prized 6.5 digit meter may well perform significantly worse than another much cheaper device with only 4.5 digits. It is very hard to find meters even amongst the elite that prioritise ac accuracy. The poor ac accuracy performance is missed in so many reviews so we should make a point of discussing ac accuracy and especially for small signals..
...

I had saved an old 1980's Fluke 8506A thermal RMS meter from scrap and repaired it.  Once working, I used my HP34401A as a reference to align the DC stages.  See the following thread for background:

https://www.eevblog.com/forum/testgear/some-old-school-instruments-showing-how-its-done-(hp-3325a-and-fluke-8506a)/

When it came to the AC alignment, I left it alone because of the reason you mention.  The accuracy and MHz bandwidth specs are nothing I could match.  The AC specs for this meter may be found in the manual. 

https://pearl-hifi.com/06_Lit_Archive/15_Mfrs_Publications/15_Fluke/Fluke_8506A.pdf

Note the drift in table 1-3.   

I think if I wanted to verify the cheap handheld meters AC performance, I would start by having this meter aligned and calibrated.   I had looked for source at the time I was doing the repairs but didn't find anyone I would trust and don't believe Fluke would still support it.   I would guess the cost for the work,  including a report would have exceeded what I would be willing to pay.  It's all pots, then the time for things to stabilize. If they told be a >$1000 USD, I wouldn't be too surprised.

[bdunham7]

I think if I wanted to verify the cheap handheld meters AC performance, I would start by having this meter aligned and calibrated.   I had looked for source at the time I was doing the repairs but didn't find anyone I would trust and don't believe Fluke would still support it.   I would guess the cost for the work,  including a report would have exceeded what I would be willing to pay.  It's all pots, then the time for things to stabilize. If they told be a >$1000 USD, I wouldn't be too surprised.

Last I checked Fluke would calibrate it (no repair service offered however) and the cost was indeed more than $1000.   The proble with 'regular' cal labs is I think they would only do the front-panel software calibration and consider any knob-twiddling as "repair".  I'm putzing with mine right now as I'll soon have the opportunity to compare it to a 3458A and 5200A.  Wish me luck.  But I'm not sure you'd actually need to go to these extremes, at least for any handhelds I've seen.  For extreme low-level testing, I think an in-cal 34401A and a ratio transformer would suffice?

[joeqsmith]

Good to hear you are still working on yours.  I gave up trying to reverse engineer their firmware and shelfed the meter.  I have thought about taking it to work and comparing it with a calibrated 3458A.  Maybe I could then go through the AC alignment.   My problem would be coming up with a suitable source.  Post how it works out for you. 

Hard to say when it comes to reviewing the cheap meters.  I don't watch very many reviews and see little value in the typical unboxing, connect them to a battery and resistor then giving them 5 stars.  They serve a purpose for people who love to have their purchasing decisions validated.  I liked Dave's reviews because of the humor.   

The only reason I make any measurements is to show that a meter has changed (normally damaged), not how accurate it is or that it meets the manufactures claims.   I may make the same measurements 10 or more times during a video looking for that one time where it shifts, signaling a problem.   

[joeqsmith]

Specs for some of the left over hand helds that I could still test.  Sorry Keysight, your meter didn't come close to surviving and could not be repaired.  The 121GW was saved by the fact I bought two of them and never exposed one of them to any potential destructive tests.  The crappo TPI required a new controller IC.  The 181A required a new power supply IC.  The Yokagowa required several ICs but all were common.  The BM869s required new transistors for the high speed clamp.   Fluke 189, tested a beat up used one but no idea what it would take to damage it.  Typical of Flukes later products.

Nothing comes even close to matching specs of a high end meter or even the Fluke thermal RMS relic previously mentioned.  If calibrated, I could certainly use it to show the accuracy of these low cost meters.* 


*The Ultra/Prime what ever Gossen calls it now, costs over $1100 USD today.  But consider that old Fluke was over $10,000 back in the early 80s when it was made.  It's all relative. 

[bdunham7]

Post how it works out for you.

Well, surprisingly enough the 8506A is fixed and now appears to be within specs as close as can be determined using a 3458A, a 5200A, a ratio transformer and other assorted instruments. 

However, it also turns out to be not very suitable for testing low-level AC peformance of handhelds, at least not directly.  One unfortunate aspect of the Thermal RMS sensor is that it doesn't perform well at the low end of the scale and this is way worse than a TRMS chip would be.  It has reduced accuracy below 25% of full scale and is not specified below 10%.  This is apparently why they have the .100/.300/1/3/10/30 etc ranges--to keep the sensor above the 25% level, at least when autoranging.  The meter is not specified for reading anything below 12.5mVAC.  According to the explanation in the manual, the excessive residual counts error below the specified minimums are due to 1/f noise, but the meter also picks up all sorts of RF and line-conducted noise.  With a LSD of 1µV and a BW >1MHz, this isn't too surprising as an AM radio station can be under 1MHz and provide 100µV or more with even a poor antenna.  I had to rearrange my bench, eliminate my UPS and turn a bunch of stuff off to even get close to proper low-level performance.

But now perhaps I can try some meters.  I'm open to suggestions as to what the test standards should be for low-level performance.  I was going to start at 1kHz and determine the minimum voltage for a meter to be accurate to within 1% or 1 count, whichever is greater.  Or since the OP wanted to look at audio, I could also look at 20Hz and 20kHz, both of which are performance verification test points for the 8506A.

[Kleinstein]

For a low level test signal one could use a voltage divider (resistors or transformer). So the source and ref. meter would be at a higher voltage, like 100x the small signal. This way one would not have an issue with rather small values.  For the moderately low frequencies a DDS based signal generator should be quite stable and flat in the frequency response, expecially if there is no AC coupling at the low end.

The low amplitude performance may scatter between units, depending on the internal offsets at the analog RMS chip and how they try to compensate (at least to some approximation) for the meters own noise.

At the lower frequency end, like 20 Hz one can expect some deviations with different waveforms. So it may be worth looking at a sine and square wave (maybe also half wave rectifier from an arb generator)  for the low frequencies. I would expect some deviation at low frequencies, especially with meters that have a relatively fast response and thus not so slow filter at the RMS chip.

Good thermal converters are still the standard, but not all thermal converters are that good. A good meter with digital RMS (e.g. KS 3446x or Keithley DMM6500) may be the more practical reference meter for a small AC signal, at least for the lower frequencies. At least in principle they can compensate for the own noise and get good performance also at a small fraction of the range. The thermal converters are more thing for higher frequencies, not so much small signals.

[joeqsmith]

Post how it works out for you.

Well, surprisingly enough the 8506A is fixed and now appears to be within specs as close as can be determined using a 3458A, a 5200A, a ratio transformer and other assorted instruments. 
...
But now perhaps I can try some meters.  I'm open to suggestions as to what the test standards should be for low-level performance.  I was going to start at 1kHz and determine the minimum voltage for a meter to be accurate to within 1% or 1 count, whichever is greater.  Or since the OP wanted to look at audio, I could also look at 20Hz and 20kHz, both of which are performance verification test points for the 8506A.

Good to hear you have it sorted out.   Looks like Richard lost interest.   Odd as they were the ones asking. 

[Svgeesus]

The perspective is that for most users, small signal AC performance is not that important. As I said, for some like yourself it may be THE spec you are after, but for most people it's just not, and that's why reviews have evolved to not even bother measuring it. Anyone with experience with RMS converter chips knows that small signal perfornace is usually poor, the bandwidth drops and so can the accuracy, so you avoid doing that if possible.

And for others, audio bandwidth and microvolt measurements is the primary use case and frankly we couldn't care less about high voltage safety because we are measuring on ±15V systems (what the op-amp providers now call "high voltage") all the time; if we have some need to probe a 400V system we will call in a professional, who will have their own Fluke or whatever.

Not that we would expect a review to omit that - but we would likely skip over that part.

[bdunham7]

And for others, audio bandwidth and microvolt measurements is the primary use case and frankly we couldn't care less about high voltage safety because we are measuring on ±15V systems (what the op-amp providers now call "high voltage") all the time

OK, then what levels, BW and accuracy are you looking for?  µVAC is pretty tough for a regular DMM and even tougher if you want TRMS.  I don't have many handheld DMMs to test, but I can try what I have--but I'm not sure what parameters would be most important. 

[bdunham7]

Good to hear you have it sorted out.   Looks like Richard lost interest.   Odd as they were the ones asking.

Perhaps, or perhaps just busy with the holidays.  In any case, I found a relevant white paper courtesy of the U.S. Army.

https://www.govinfo.gov/content/pkg/GOVPUB-C13-83cb89e6828c42f3920d25f1c0c1001c/pdf/GOVPUB-C13-83cb89e6828c42f3920d25f1c0c1001c.pdf

[miro123]

Accuracy
Range
Bandwidth
RMS
Decibels

The short answer is look at
https://www.youtube.com/@N8FDY
If you prefer text review use the github link
https://github.com/TomWilkinson/Multimeter-Reviews/tree/main
Take all those reviews with grain of salt if you are interested in accuracy. The most multimeters reviews are kind of "one night stand ".  Evaluation requires to use single multimeters for long time. Unfortunately this group of people are not youtubers nor web posters. 
 

[joeqsmith]

Good to hear you have it sorted out.   Looks like Richard lost interest.   Odd as they were the ones asking.

Perhaps, or perhaps just busy with the holidays.  In any case, I found a relevant white paper courtesy of the U.S. Army.

https://www.govinfo.gov/content/pkg/GOVPUB-C13-83cb89e6828c42f3920d25f1c0c1001c/pdf/GOVPUB-C13-83cb89e6828c42f3920d25f1c0c1001c.pdf

Perhaps.  I wouldn't put in a lot of effort unless it is something you are personally interested in.   Thanks for the link.   

[bdunham7]

The short answer is look at
https://www.youtube.com/@N8FDY
If you prefer text review use the github link
https://github.com/TomWilkinson/Multimeter-Reviews/tree/main
Take all those reviews with grain of salt if you are interested in accuracy.

A grain of salt indeed.  Tom certainly has a wide assortment of multimeters and I have to applaud him for providing the written reviews as well as his straightforward approach.  There are, however, some errors interpreting specifications and he also comes to the conclusion that certain models do not meet their specifications, which I think may be erroneous.  The Fluke 289, Fluke 177/179, Greenlee 860A and UNI-T 181A reviews may potentially have these errors, and perhaps others since I didn't read them all.  However, even these questionable conclusions raise some substantial issues, just not the ones that they appear to be at first glance. 

The only one I actually have is the 289, so naturally I looked at that one and saw that he claims that it does not meet its DC accuracy specifications at 1mVDC and at various points from ~300-500VDC.  The error at 1mVDC is because the 289 has a 50mVDC range with an additional x10 gain stage and the specifications state (in the fine print) that the tolerances for this range are stated after nulling any offset--IOW you have to zero it first.  He states that he does not have reference standards but instead relies on unidentified power sources and a Keithley DMM6500 in parallel as a reference.  He has a picture comparing a number of meters reading a source in parallel, all showing noticeable error with some in-spec and some out depending on the particular meter and its tolerances.  I can think of one possible cause for this (other than the meters being out of spec), but I'll wait to state until others can make some suggestions.

[guenthert]

The short answer is look at
https://www.youtube.com/@N8FDY
If you prefer text review use the github link
https://github.com/TomWilkinson/Multimeter-Reviews/tree/main
Take all those reviews with grain of salt if you are interested in accuracy.

A grain of salt indeed.  Tom certainly has a wide assortment of multimeters and I have to applaud him for providing the written reviews as well as his straightforward approach.  There are, however, some errors interpreting specifications and he also comes to the conclusion that certain models do not meet their specifications, which I think may be erroneous.  The Fluke 289, Fluke 177/179, Greenlee 860A and UNI-T 181A reviews may potentially have these errors, and perhaps others since I didn't read them all.  However, even these questionable conclusions raise some substantial issues, just not the ones that they appear to be at first glance. 

The only one I actually have is the 289, so naturally I looked at that one and saw that he claims that it does not meet its DC accuracy specifications at 1mVDC and at various points from ~300-500VDC.  The error at 1mVDC is because the 289 has a 50mVDC range with an additional x10 gain stage and the specifications state (in the fine print) that the tolerances for this range are stated after nulling any offset--IOW you have to zero it first.  He states that he does not have reference standards but instead relies on unidentified power sources and a Keithley DMM6500 in parallel as a reference.  He has a picture comparing a number of meters reading a source in parallel, all showing noticeable error with some in-spec and some out depending on the particular meter and its tolerances.  I can think of one possible cause for this (other than the meters being out of spec), but I'll wait to state until others can make some suggestions.

I'm always uncomfortable measuring a source with multiple meters in parallel / simultaneously.  If the source is not of very low impedance, the bias current (might average to a few pA, but easily spikes in the nA range) rakes havoc on those measurements.  For handheld DMMs, it's probably more the limited (and not all that well defined) input resistance which gets in the way.

Regarding True RMS measurements: in the early 70s, -hp- published a pamphlet warning of the traps of average converting meters: http://hparchive.com/Application_Notes/HP-AN-124.pdf

For the Datron/Wavetek 1271 (I suppose also for the 1281) there is a True RMS option board, which does not use a thermal converter (because, as they claim, that would be too slow); instead they use a hairy assembly of squaring, log- and antilog circuits (the instrument is old enough that the digital circuitry was too slow for the sampling approach).  The claim good accuracy down to 1%FS with 100mV the lowest range.

[Kleinstein]

A parallel connection of voltmeters should not be an issue. The bias and loading from normal impedance, like 10 M or even lower with analog meters is effecting all the meters in the same way and is thus no issule - it would only be a problem if the cables are rather high resistance - so no real issue with less than 9 digits.

The switching glitches may be an issue with some high end bench meters at high resolution. The lower resolution meters usually have input protection with enough resistance to filter out the spikes and both ends. A good meter should also be somewhat tolerant to some spikes / variations in the signal. If the parallel connection causes problems there should be 2 meters at fault: one causing interference and one being a bit sensitive.

The RMS circuit in the Datron meters is kind of the old form (maybe a little better) of what is more commonly found inside the RMS-DC onverter chips, like the AD637 and similar and still used in quite some meters. There is nothing really wrong with this. Thermal RMS is old style and slow and thus no longer used from practical multimeters. It is still a thing for high frequencies and maybe the metrology side.  The digital sampling method is getting increasingly popular, but not necessary better (especially not in the common low power versions found in handheld DMMs) - it has different pros and cons. So it helps to know which method is used.

[2N3055]

I'm always uncomfortable measuring a source with multiple meters in parallel / simultaneously.  If the source is not of very low impedance, the bias current (might average to a few pA, but easily spikes in the nA range) rakes havoc on those measurements.  For handheld DMMs, it's probably more the limited (and not all that well defined) input resistance which gets in the way.

Loading the source will change absolute value, but you still can see accurate relative comparison to reference voltmeter.

With AC measurements, it usually creates more problem that most meters have no well defined complex impedance, mainly input capacitance, and fact that it changes all the time between the ranges.
Capacitances will easily be in hundreds of pF and will be considerable load...

Both MTX3293 (260/68 pF) and BM869S (98/75pF) change capacitance between low and high range.
DM3068 (155pF) is very stable. It does not change through the ranges..

As for resistive part of input impedance, that is well defined in datasheets on all meters I have.

But as stated before, multimeters are not made to be excellent wideband AC voltmeters.
They are Swiss Army knifes of the T&M industry.

[David Hess]

The switching glitches may be an issue with some high end bench meters at high resolution. The lower resolution meters usually have input protection with enough resistance to filter out the spikes and both ends. A good meter should also be somewhat tolerant to some spikes / variations in the signal. If the parallel connection causes problems there should be 2 meters at fault: one causing interference and one being a bit sensitive.

Some combinations of meters, even lower resolution ones, will have problems with charge injection from the automatic zero or chopper cycle.  Usually I see this when measuring the input resistance of one meter using another, but it also shows up with high impedance sources.

I keep a couple of old multimeters which do not produce charge injection for sanity measurements.  They use brute force with precision matched JFET input stages instead of automatic zero or chopping.

As for resistive part of input impedance, that is well defined in datasheets on all meters I have.

I would be really careful about that, especially with handheld automatic ranging meters.  They often specify 1 or 10 megohm input resistance even when it varies significantly because of how the switching for the input divider is configured.  They use a divider where different taps are grounded, which causes the input resistance to change.  Moderate source resistances, or a high voltage probe, will then allow the reading to change on different ranges by an amount greater than the accuracy rating of the meter.

Most datasheets just say 1 or 10 megohms without specifying this.  The alternative constant input resistance meters (1) sometimes specify 10 megohms with a tolerance like 1% or 0.1% which is much tighter than the variable input resistance case which is more like 10%.

This post shows what is going on with variable input resistance decade dividers on autoranging multimeters.

(1) Bench multimeters or manual ranging handheld multimeters, with some exceptions.  I have yet to find a handheld automatic ranging multimeter which has constant input resistance, but I assume they exist.  It is practically impossible to tell without measuring it.

[2N3055]

As for resistive part of input impedance, that is well defined in datasheets on all meters I have.

I would be really careful about that, especially with handheld automatic ranging meters.  They often specify 1 or 10 megohm input resistance even when it varies significantly because of how the switching for the input divider is configured.  They use a divider where different taps are grounded, which causes the input resistance to change.  Moderate source resistances, or a high voltage probe, will then allow the reading to change on different ranges by an amount greater than the accuracy rating of the meter.

Most datasheets just say 1 or 10 megohms without specifying this.  The alternative constant input resistance meters (1) sometimes specify 10 megohms with a tolerance like 1% or 0.1% which is much tighter than the variable input resistance case which is more like 10%.

This post shows what is going on with variable input resistance decade dividers on autoranging multimeters.

(1) Bench multimeters or manual ranging handheld multimeters, with some exceptions.  I have yet to find a handheld automatic ranging multimeter which has constant input resistance, but I assume they exist.  It is practically impossible to tell without measuring it.

Thank you for your good intentions and good advice. But I'm careful.

I don't do Kaiweets or something like that.
I have Metrix, Brymen, Fluke and Rigol.
They all have detailed datasheets and yes, I'm like you and verified it myself.

If you noticed, I said that ohmic resistances are well defined (and verified to be true) but capacitance does vary and have provided a measured data.

But I think we might have a slight disconnect here.
I'm not saying all these meters necessarily have single, unchanging ohmic resistance. It changes but it is well documented and verified to be such.
MTX 3293 even shows it right there on the screen as you go through ranges when on DC :

100mV (Imp 10MΩ)
1000mV (Imp 11MΩ)
10V (Imp 10.5MΩ)
100V (Imp 10.05MΩ)
1000V (Imp 10MΩ)

OTOH  DM3068 is rock solid 10,026MΩ 200m/2/20V and 10,014MΩ 200/1000V

[David Hess]

I don't do Kaiweets or something like that.
I have Metrix, Brymen, Fluke and Rigol.

...

I'm not saying all these meters necessarily have single, unchanging ohmic resistance. It changes but it is well documented and verified to be such.

Fluke is one of the manufacturers which does not accurately document the input resistance of their meters.  They have an application note which is included with their high voltage probes which has an incomplete list of which of their meters have compromised accuracy because their input resistance changes.  They suggest applying a correction factor.  I know Brymen does the same thing.  I searched their entire line of multimeters for one with a constant 10 megohm input resistance and as far as I can tell, they have none.

All of my high end autoranging handheld multimeters specify 10 megohms, but their input resistance changes as described in the post that I linked.

B&K autoranging handheld multimeters sometimes have detailed specifications with different resistance specified for different ranges, and sometimes simply say 10 megohms.  Do the later ones have a constant input resistance?  I have no idea, but at least they seem to be distinguishing the two unlike Fluke and Brymen.

[bdunham7]

I'm always uncomfortable measuring a source with multiple meters in parallel / simultaneously.  If the source is not of very low impedance, the bias current (might average to a few pA, but easily spikes in the nA range) rakes havoc on those measurements.  For handheld DMMs, it's probably more the limited (and not all that well defined) input resistance which gets in the way.

There can be issues when using meters in parallel, but usually it is the best of several imperfect choices.  The problem as I see it is that even when in parallel, a disagreement between meters doesn't necessarily result only from calibration issues.  I don't know what happened in the tests that are mentioned above, but I do know that there is a huge difference in the normal mode rejection spec for the Fluke 289 (which has been deemed "out of spec" by the reviewer) and the Keithley DMM6500.  If everything is set up optimally, the stated values are 60dB for the 289, 100dB for the DMM6500, both at 60Hz respectively.  The reviewer makes no mention of the voltage source other than to say that he doesn't have reference standards.

There are many other experimental errors that can crop up using a parallel meter as a reference, but most of those don't really apply to measuring 490V with a handheld DMM.  Differing responses to common or normal mode AC or noise is about the only thing I can think of.