Meter accuracy

[torch]

I am currently refurbishing my tube tester (a Tripplet 3444). One important step is to set the line voltage measurement and the manual specifies that the meter used to do this must have an AC accuracy of 1/2% or better. My 'good' meter (a Goldstar 334, which was a $200 meter 20 years ago) has a claimed AC voltage accuracy of 1% -- when it was new. So maybe it's time to buy a new meter. I've looked at the claimed accuracy of some of the units in Dave's $100 meter shootout, the best seems to be 0.8% -- close but no cigar.

This Tripplet is almost as old as me; how much meter do I have to buy to get 1960s era accuracy? Is there a simple and reliable way to check the accuracy of my meter that doesn't involve using the same meter to measure the value of a resistor & current?

[EEVblog]

Now you get into the murky subject of True RMS.
On most True RMS meters, 0.5% AC accuracy would be very good indeed.
Even the Fluke 87V is only 0.7% with True RMS.
But the kicker is the cheaper non-true rms average responding Fluke 83V is 0.5% because it's easier to get better accuracy when you don't use a True RMS converter chip.
So if you know your signal is precisely a sine wave, then an averaging responding meter will give you better accuracy/$ here.

Dave.

[Simon]

essentially to calibrate a measuring instrument you need a reference that is at least an order of magnitude if not two better than the one being calibrated - just a fact of measuring life 

[torch]

But the kicker is the cheaper non-true rms average responding Fluke 83V is 0.5% because it's easier to get better accuracy when you don't use a True RMS converter chip.

Ok, now that surprises the #&!! out of me. 

essentially to calibrate a measuring instrument you need a reference that is at least an order of magnitude if not two better than the one being calibrated - just a fact of measuring life

That doesn't.     (it's the same in machining -- if you need to hold a 1/2 thou tolerance, you need a mike that's accurate to a tenth.)  Oh well, maybe I'll just use the $15 el cheapo I carry in the car -- if cheaper is better, that one should be bang on! 

[Simon]

 (it's the same in machining -- if you need to hold a 1/2 thou tolerance, you need a mike that's accurate to a tenth.) 

Yes I'm a quality inspector for a so called engineering company although they have no given me formal training so I go mostly by intuition (better that some peoples theories  )

[ejeffrey]

This is a perfect job for a bench meter.  The basic fluke, keithly, and agilent bench meters all have 0.2% accuracy or better on ACV.

On the other hand, I wouldn't put any stock at all in that accuracy requirement -- it is an ill posed measurement problem.  Line voltage is simply not stable to 0.5% over either short or long time scales.  Measure with a 1% meter and you will be fine.

IIRC, the primary temperature dependence of tube equipment is from the change in the cathode heater current since the power goes as V^2.  At least some people who refurbish valve guitar amps put regulated supplies on the cathode heater -- primarily to avoid early burn-out due to the higher line voltages now common without rewinding the transformer.  If you are worried you might look in to separately regulating any sensitive voltages.

[torch]

It is regulated, and this is a key adjustment that affects the accuracy of the entire unit. They stress the importance rather firmly:

8. Use an accurate AC voltmeter (±1/2% or better ) with sensitivity of at least 5,000 ohms per volt and connect across pins # 1 and #2 in parallel with an 18 ohm, 5 watt load.

9. Note meter reading while adjusting the LINE VOLTAGE control for LINE TEST.

10. If at 6 .3 volts, the alignment of the LINE TEST is off, no attempt should be made to change the calibration resistor R4 unless absolute certainty of the accuracy of the voltmeter is known to be ± 1/2% at 6.3 volts.

NOTE- Due to the importance of this adjustment. It is recommended that it be made only at the factory or Field Service Station.

Unfortunately there seems to be a dearth of Tripplet Field Service Stations around here.

Thanks for pointing me in the right direction, although I suspect this is going to hurt. Hey Dave, done any shootouts on bench meters lately?

[torch]

Update: IF my shiny new Hantek DDS3X35 is actually putting out the same voltage it's told to (and this is by no means certain -- there's nothing in the spec about that) then the reading on my 20 yr old meter is 1.7% low (@60hz). And the el cheapo meter I carry in the car is so far out it isn't in the same ballpark. Well, so much for that idea...

[saturation]

If you don't have a standard, anything around you may be better than have no calibration.

If your old meters haven't been calibrated but neither have they been repaired, they are likely closer to a real value than the meter you've just repaired.  At least it will be better than having nothing at all.

As for the Hantek 3x25 and if you have a oscilloscope, at least it won't be more off that 4% measured with the DSO's voltmeter function.  If it can give you a good Vpp, you can convert it to RMS from there.

[torch]

Anybody ever used (or even heard of!) a Vichy VC8145? Lot's of features. Too bad it's not battery powered but the claimed specs seem pretty decent (ok, not Fluke bench meter level, but at least it meets .05% ACV), and the price isn't too bad either (eg: http://cgi.ebay.com/170642113428, but there's several of other vendors dealing them too). Is this a potential bench meter version of the Rigol scope (ie: "good enough for the rest of us")?

[vk6zgo]

I am very dubious about the necessity for that degree of accuracy in a tube tester.
After all,these things are usually used in a "go-no go" capacity,rather than to actually  critically determine tube characteristics.

Have you changed all the high value resistors in the device?
If not,they will probably have drifted,& will have more effect on your accuracy than the AC voltage reading.
Check with your existing meter & if it reads close,leave it.

The only meters in common use when these things were made that approached that degree of accuracy,were
called "Sub-Standard Voltmeters,& were  in a wooden case about 1ft square (300mm x300mm).

We used to use these devices to adjust filament volts on Transmitters.
Even so, when DMMs appeared,they were retired,& we accepted the DMMs accuracy as OK for this purpose.

VK6ZGO

[torch]

I am very dubious about the necessity for that degree of accuracy in a tube tester.
After all,these things are usually used in a "go-no go" capacity,rather than to actually  critically determine tube characteristics.

The Tripplet 3444 is a bit more than most "go-no go" testers, although it could of course be used in that capacity too: http://www.indianaradios.com/Triplett%203444%20Tube%20Tester.htm. Tripplet called it a "Laboratory Tube Analyzer" which might have been a little bit of marketing hype, but it is definitely considered one of the better models out there, on par with the best portables by Hickock. In proper fighting trim, this one can be used to determine tube characteristics, which can be important for finding balanced pairs eg: audio amps. Do I personally, absolutely need that accuracy -- not at the moment. But since I still have the damn thing I thought it would be nice to bring it back to spec.

Have you changed all the high value resistors in the device?
If not,they will probably have drifted,& will have more effect on your accuracy than the AC voltage reading.

They're done and replacement will be part of this exercise. I'm still hemming and hawing over re-tubing it. It's a bit of a catch-22: How do you check the tubes in the tube tester without access to another tube tester?

My current DMM was considered pretty good for it's day, and certainly met my needs "in the field", but it hasn't had an easy life initially bouncing around in the back of a truck and later in my garage. It's never been calibrated or adjusted. Dave points out that a second meter is pretty useful to have sometimes. Rather than buy a cheaper meter as a backup, my thought is I would be better to buy a decent meter if I'm serious about getting more into electronics and relegate "old faithful" to back-up status. No?

[ejeffrey]

It is regulated, and this is a key adjustment that affects the accuracy of the entire unit. They stress the importance rather firmly:

Maybe then I am not understanding what is actually being adjusted.  This sounds like you have a power transformer that has a nominal 6.3 volt output winding, but due to variations in the mains line voltage will be higher or lower.  Then, by adjusting either the transformer tap for coarse tuning or a trim resistor for fine tuning you are trying to get it to exactly 6.3 VAC +/- 0.5%.   If that is what is happening, it can't possibly work as the output is ratiometric to the line voltage which varies by more than 0.5% on time scales from minutes to hours, months and years.  On the other hand, if the supply is genuinely regulated with feedback and you are trimming the feedback resistors, that all sounds fine.

What is the 6.3 VAC supply?  I guess it must be the grid test signal then?

[/quote]
Thanks for pointing me in the right direction, although I suspect this is going to hurt. Hey Dave, done any shootouts on bench meters lately? [/quote]

I can save you and dave the time.  The big name brands are fluke, keithley, and agilent.  Stay away from the keithley, they have great specs and are cheap (6.5 digit for the price of 5.5 digit from the others) but they are unreliably pieces of crap and tend to die 3 months out of warranty.  I replaced my dead keithley 2000 series with a fluke and like it so far.  Agilent is getting back into this business, I know some people who use them, but I don't have anything to say about them.  If you can find an old school HP, they are the best of the bunch, but the calibration may have drifted with age so you may be in a chicken / egg situation for this particular task.  I don't know about other brands -- bench meters are a more specialized market, so there is less variety in both features and price than hand held meters.  If you know someone in a physics lab you can probably borrow a meter for an afternoon...

[alm]

I can save you and dave the time.  The big name brands are fluke, keithley, and agilent.  Stay away from the keithley, they have great specs and are cheap (6.5 digit for the price of 5.5 digit from the others) but they are unreliably pieces of crap and tend to die 3 months out of warranty.  I replaced my dead keithley 2000 series with a fluke and like it so far.

The firmware of the recent(ish) Keithley 2100 appears to be crap, no idea about reliability. Not aware of any complaints about the 20xx series. How large is the sample size you used to draw conclusions about their reliability? I see plenty in use, so it can't be that bad.

Agilent is getting back into this business, I know some people who use them, but I don't have anything to say about them. 

Getting back? The Agilent 34401A is the industry standard (being replaced by the 34410A), and has been in continuous production since early nineties or so. Agilent has been producing them since they where spun off HP, and HP has been making bench meters since at least the sixties. Fluke has actually been getting back (didn't have a 5.5/6.5 digit meter for a significant number of years), and I've heard some complaints about the durability of their split banana jacks they use to implement their '2x4W' feature. Experiences I've heard about the 8846A weren't that positive.

If you can find an old school HP, they are the best of the bunch

What's the difference between a HP 34401A and an Agilent 34401A, except that the former has had more time to drift out of calibration?

I don't know about other brands -- bench meters are a more specialized market, so there is less variety in both features and price than hand held meters.  If you know someone in a physics lab you can probably borrow a meter for an afternoon...

Current high-precision products it's just Agilent, Fluke and Keithley as far as I know (you can also get the Fluke meters with a Tektronix logo ). If you include used products, manufacturers that produced them within say the last twenty years or so include Datron/Wavetek, Prema, Schlumberger and Solartron.

[tekfan]

If you can find an old school HP, they are the best of the bunch

What's the difference between a HP 34401A and an Agilent 34401A, except that the former has had more time to drift out of calibration?

One could argue that an older meter had time to age and stabilize, so calibrating an old meter can actually get you better results in tems of voltage reference drift.
High precision quartz crystals are actually aged to get the required drift specs. I would imagine that something similar can be applied to voltage references.


As for the true RMS thing, if you really want a high precision RMS reading, you can get an analog meter with a big scale. AVO meters have really big meter movements and can potentially provide better than 0.5% accuracy. The instrument of course must be calibrated to an AC source with a known and stable amplitude.


Well even the Datron 1281 doesn't have a very precise RMS converter. This multimeter has basically been a standard in almost every calibration lab. DC accuracy better than 0.0006% but AC only 1%.
http://www.eurekaspot.com/sp.cfm/DMMDVM/WAVDAT/1281.html

[alm]

One could argue that an older meter had time to age and stabilize, so calibrating an old meter can actually get you better results in tems of voltage reference drift.
High precision quartz crystals are actually aged to get the required drift specs. I would imagine that something similar can be applied to voltage references.

That's also the case for voltage references. I believe manufactures like Agilent tend to age the references for optimum stability (or at least they used to in the HP days, maybe the process has improved enough to make this obsolete). One example is the high-stability option on the Agilent 3458A, which is just a reference aged for some extra time and monitored to prove it's got low drift. So if the old equipment was recently calibrated, it's likely to be better.

As for the true RMS thing, if you really want a high precision RMS reading, you can get an analog meter with a big scale. AVO meters have really big meter movements and can potentially provide better than 0.5% accuracy. The instrument of course must be calibrated to an AC source with a known and stable amplitude.

Do they measure RMS? How about other sources of inaccuracy than the meter movement? The fact that the dial may be accurate to 0.5% does not make the measurement that accurate, just like a 4 digit LCD does not make a 0.1% accurate DMM.

Well even the Datron 1281 doesn't have a very precise RMS converter. This multimeter has basically been a standard in almost every calibration lab. DC accuracy better than 0.0006% but AC only 1%.
http://www.eurekaspot.com/sp.cfm/DMMDVM/WAVDAT/1281.html

The Agilent 3458A would do it I think, and so would Fluke's high-end offering (based on the Datron technology), but something tells me that's slightly overkill for this task. They probably used a high-end average-responding VOM in those days. Is it possible that 1/2% means 1 or 2% (would seem like a strange way to specify accuracy)?

[vk6zgo]

I doubt very much if the 6.3volt AC supply was regulated. Regulation of an AC supply was extremely difficult  back then,& still is, if it comes to that!

It sounds to me like , in use,this tester has a control called something like "AC Set".
Using this,you can adjust the AC to a "set" line ,then do your tests knowing that you have compensated for mains variations.
The AVO Valve Characteristic Tester has a similar feature.

The meter is factory set so that it lines up with this "set" line when the 6.3volt supply is correct.This is the adjustment you are duplicating.
You basically need to do this adjustment,while checking back to see if the 6.3 volt reading has varied,as it well may.

The Sub-Standard  meters I referred to earlier, had a calibration chart included,so corrections could be made for
the individual meter.
Maybe, if you are really very concerned ,you could find someone with a meter of higher accuracy,test your meter along with the other one at voltages around 6.3volts AC,& make up a calibration chart in this manner--or just borrow the more accurate meter!

VK6ZGO

[torch]

Maybe then I am not understanding what is actually being adjusted.  This sounds like you have a power transformer that has a nominal 6.3 volt output winding, but due to variations in the mains line voltage will be higher or lower.  Then, by adjusting either the transformer tap for coarse tuning or a trim resistor for fine tuning you are trying to get it to exactly 6.3 VAC +/- 0.5%.   If that is what is happening, it can't possibly work as the output is ratiometric to the line voltage which varies by more than 0.5% on time scales from minutes to hours, months and years.  On the other hand, if the supply is genuinely regulated with feedback and you are trimming the feedback resistors, that all sounds fine.

What is the 6.3 VAC supply?  I guess it must be the grid test signal then?

The load is connected to the filament voltage supply. The supply is described as "Twenty-three AC voltages, closely regulated". The circuit diagram (included in the manual at the link I posted earlier) is rather confusing the way it's drawn -- at least to me. Some of the rotary switches are drawn one way, others another. The wiring is also hard to follow -- no circuit boards, it's all handwired with components hanging off switches, etc. The original "dead bug".

But from what I can see, the filament switch is connected to the main (33 tap) transformer, resistor network & 1/2 wave rectifier. That is then regulated -- I think -- through a tube-driven circuit with negative feedback through a full wave bridge rectifier. The "Calibration resistor R4" connects the filament switch + to one side of the oscillator transformer through some resistors, the amplifier tube and the meter when the "Line Test" switch is pressed.

Gross adjustment of the line voltage before each use is through a variable pot on the primary side of the transformer -- 1 leg of the supply is connected to the wiper and the ends are connected between the centre tap and one end of the primary, the other leg to the opposite end. Cute.

[torch]

It sounds to me like , in use,this tester has a control called something like "AC Set". Using this,you can adjust the AC to a "set" line ,then do your tests knowing that you have compensated for mains variations.

The AVO Valve Characteristic Tester has a similar feature.

The meter is factory set so that it lines up with this "set" line when the 6.3volt supply is correct.This is the adjustment you are duplicating. You basically need to do this adjustment,while checking back to see if the 6.3 volt reading has varied,as it well may.

Yes, sounds like you are describing what Triplett refers to as the "Line Test" switch.

Maybe, if you are really very concerned ,you could find someone with a meter of higher accuracy,test your meter along with the other one at voltages around 6.3volts AC,& make up a calibration chart in this manner--or just borrow the more accurate meter!

I wish that it were so! Unfortunately I live outside a small town in northern Ontario. Most of my acquaintances sell junk to tourists, cut timber or peddle real estate -- not much in the way of industry or technology here. A few of my neighbours know what a meter looks like. A significant portion of those are familiar with what a smoking meter looks like, the others know enough not to check the mains for power using the AC current setting. The remainder want to know what the extra wire is for on a 3-way light switch.

Ok, maybe I'm exaggerating a tiny bit.  But not much.

[vk6zgo]

Yeah, I had a look at the circuit,& from memory,it looks a lot like the AVO one.

The 12AU7 is for the "Electronic Milliammeter",& the 6C4 is for the 5kHz test signal.

There is nothing in the circuit that comes "within a bulls roar" of being a regulator for the AC filament voltages.
I think the "Closely regulated" stuff is hyperbole by Hickock's Sales Dept.
They probably mean that,when the Line voltage is set correctly,the transformer secondaries produce voltages very close'
to those quoted.In other words,a very accurate transformer.
This,of course, is why they have to adjust the primary voltage,using the set voltage control.

The apparent anomaly with reading accuracy between  "True RMS" & "non-True RMS" is due to the fact that the latter read Average voltage very accurately,using a  bridge rectifier just as per the old moving coil meters.
They then have a correction factor added to the reading to read RMS --For sinewaves only!

The present generation of "True RMS" meters are slightly less accurate for sinewaves,but they will read RMS for most other waveshapes with the same accuracy.

Re: getting access to a more accurate meter. Is there a Radio Broadcast  Station in the area ?
I know that the Transmitters aren't manned anymore,but possibly there is a regional studio,& if you talk nicely to them,you
might get access to a meter of higher accuracy.

Another possibility is to find a Radio Ham in the area.You could ask on QRZ.com.

Or, just use the existing meter---Or,assume the line Test metering is still correct,which is what most users probably did!

VK6ZGO

[tekfan]

As for the true RMS thing, if you really want a high precision RMS reading, you can get an analog meter with a big scale. AVO meters have really big meter movements and can potentially provide better than 0.5% accuracy. The instrument of course must be calibrated to an AC source with a known and stable amplitude.

Do they measure RMS? How about other sources of inaccuracy than the meter movement? The fact that the dial may be accurate to 0.5% does not make the measurement that accurate, just like a 4 digit LCD does not make a 0.1% accurate DMM.

Yes. Almost all old analog meters with passive components measure True RMS AC. The input voltage is rectified and then applied to the meter movement. This is the same technology that is used in average responding digital meters, but the difference is that there is no filtering capacitor across the meter movement. It's a very slight modification but that's really all that is needed.

Some high frequency voltmeters also used this technique. The HF signal went trough several amplifier stages and was rectified by fast switching diodes and then fed directly to a meter movement specially designed for high frequency.

As for the meter accuracy, the most important thing in any moving coil meter is its linearity. Even if it doesn't read exactly full scale at let's say 50uA the resistors in series with the meter can be trimmed to show an exact full scale deflection.
If I remember correctly some high percision equipment from General Radio actually had more than 10 screws behind the meter to adjust its linearity.

[vk6zgo]

No,sorry, Analog moving coil meters don't read true RMS,they read Average,& the scale is modified to read RMS.
The mechanical action of the meter movement integrates the rectified 1/2 cycles from the bridge rectifier,so that the meter deflection is proportional to the average value of current.

Old-fashioned moving-iron meters read RMS directly ,if I remember correctly,but I haven't seen one of those for many years.

VK6ZGO

[tekfan]

No,sorry, Analog moving coil meters don't read true RMS,they read Average,& the scale is modified to read RMS.
The mechanical action of the meter movement integrates the rectified 1/2 cycles from the bridge rectifier,so that the meter deflection is proportional to the average value of current.

Old-fashioned moving-iron meters read RMS directly ,if I remember correctly,but I haven't seen one of those for many years.

VK6ZGO

Thanks for correcting me on that. See this is what forums are all about. You're right moving coil (d'Arnsonval) meters read the average of a rectified waveform. Moving iron vane and electrodynamic meters show RMS value directly. They also work on DC but the iron vane gets magnetized. Here the iron vane meters are still widely used in power monitoring applications where RMS values are important.

[alm]

The apparent anomaly with reading accuracy between  "True RMS" & "non-True RMS" is due to the fact that the latter read Average voltage very accurately,using a  bridge rectifier just as per the old moving coil meters.
They then have a correction factor added to the reading to read RMS --For sinewaves only!

The present generation of "True RMS" meters are slightly less accurate for sinewaves,but they will read RMS for most other waveshapes with the same accuracy.

How much distortion does it take for the average-responding meter to be less accurate than the true-RMS meter? I'm wondering if an average-responding meter doesn't just replace the problem of measuring ACV accurately to a problem of measuring distortion accurately (how else do you know it's a sine?). I wouldn't assume mains is a sine once you're talking about 0.5% accuracy with todays non-linear loads.

[vk6zgo]

I suppose one way to check the error due to distortion would to be to read the output voltage of an Audio Generator as used to test Broadcast Stations (these things have extremely low distortion),set it to 50 or 60Hz,take a measurement,then try the measurement with the same voltage level derived from the mains.

The problem here,is how do you  know it is the same level?
Perhaps you could use a Noise & Distortion meter to check the level of both signals,as they are fairly accurate.
The N&D Set would allow you to determine the distortion of the mains derived voltage anyway,so you could
maybe work backwards from there.

VK6ZGO