Incoming stuff, voltage reference..

[ErikTheNorwegian]

Sinc

[saturation]

I don't own one but they've been discussed on eevblog for some time.  zildjianboy7 is the developer, aka Doug Malone.

http://www.febo.com/pipermail/time-nuts/2008-November/034749.html

http://www.voltagestandard.com/

I've read folks comparing early Malone and Geller references on DMMs, and claim they were still reading close to when it was first obtained in ~ 2007 [ time nuts, volt nuts or hp_agilent_equipment forums.].  I think that's highly plausible on < 20,000 count DMMs, maybe even 50,000 count DMMs.  I don't hear much about using them for 6.5 digit or more and comparing their long term stability.  If built relatively well, references drift up or down in uV/year, so it could take decades before you'd see 1mV change. 

[cybergibbons]

I just got my DMMCheck from Doug. Does what it says on the tin and is good value. I think it's all a hobbyist really needs to check meters.

[MBY]

I received a pentaref some weeks ago. Me likes! But the 6 PPM/C is a little too much as my room temp seldom go under 25C. The interesting thing is that if I take the 6 PPM/C into account, the pentaref is spot on according to my calibrated Rigol DM3061 6½ digits DMM.

I'm thinking of building a "reverse oven" with a peltier to keep the pentaref at 68F (stated cal temp = 20 degrees C), but its only at the idea state. I was thinking of a peltier attached to a alu case, with the pentaref inside it, and a bigger plastic case to house everything with some thermal isolation, and then some ATTiny with some kind of temp control (maybe a good exercise in PID-control?). Ideas?

[alm]

Wouldn't it be easier to use an ovenized reference like an LM399 or LTZ1000?

[T4P]

I received a pentaref some weeks ago. Me likes! But the 6 PPM/C is a little too much as my room temp seldom go under 25C. The interesting thing is that if I take the 6 PPM/C into account, the pentaref is spot on according to my calibrated Rigol DM3061 6½ digits DMM.

I'm thinking of building a "reverse oven" with a peltier to keep the pentaref at 68F (stated cal temp = 20 degrees C), but its only at the idea state. I was thinking of a peltier attached to a alu case, with the pentaref inside it, and a bigger plastic case to house everything with some thermal isolation, and then some ATTiny with some kind of temp control (maybe a good exercise in PID-control?). Ideas?

Isn't it 5½ ?

-------------------------------------------------------------------------------
Surely would work well to tune this :
http://www.kmitl.ac.th/~kswichit/DVM7135/DVM7135.htm 
Nothing beats a 8 digit led display 

[MBY]

I received a pentaref some weeks ago. Me likes! But the 6 PPM/C is a little too much as my room temp seldom go under 25C. The interesting thing is that if I take the 6 PPM/C into account, the pentaref is spot on according to my calibrated Rigol DM3061 6½ digits DMM.

I'm thinking of building a "reverse oven" with a peltier to keep the pentaref at 68F (stated cal temp = 20 degrees C), but its only at the idea state. I was thinking of a peltier attached to a alu case, with the pentaref inside it, and a bigger plastic case to house everything with some thermal isolation, and then some ATTiny with some kind of temp control (maybe a good exercise in PID-control?). Ideas?

Isn't it 5½ ?

-------------------------------------------------------------------------------
Surely would work well to tune this :
http://www.kmitl.ac.th/~kswichit/DVM7135/DVM7135.htm 
Nothing beats a 8 digit led display 

No, 2 400 000 counts. You are probably thinking of DM3051 (I think it is named so) with 480 000 counts.

Edit: Your link shows a 2*4½ digit DMM, not a 8 digit one!

[Circuitous]

I have the DMM-Check (8 months old), Penta-Ref (4 months), and a Geller SVR (2 months).

A month ago I had tested all three references and they appeared to be very accurate.  I received a new, calibrated, Agilent 34410A DMM and decided to retest the references.  The SVR and the DMM-Check both appeared to have remained accurate, but the Penta-Ref was a different story.

The Penta-Ref had been quite accurate in February, when I checked it with a borrowed calibrated Agilent meter.  But, on 3/20 with a new 34410A it seemed off.  The next day I thought about it and checked the two 9-volt batteries which power the Penta-ref… sure enough, they were running a bit low.  With the Penta-ref switched on, the battery voltages were 8.92 and 8.97.  I replaced them with brand new batteries (9.6volts) and re-ran the tests.  That fixed it, the Penta-ref was back on track.

On a similar note, the Geller SVR provided a steady 10.0000 volts output using an input voltage as low as 10.945V.
Geller SVR 23.5° C
Value    34410A    Fluke 289
10 Volts    10.0000    10.000

Penta-Ref
           New batteries 22° C       Old batteries 23.5° C
Penta-ref     34410A    Fluke 289    34410A    Fluke 289
0.2500    0.25013    0.25013      0.25426    0.24543
0.4900    0.49007    0.49009      0.49821    0.4983
0.5100    0.51009    .51012      0.51855    0.5186
4.9000     4.9001     4.9004      4.9817     4.982
10.0000    10.0000    10.0001    10.1666    10.167

DMM-Check 23.5° C
Value    34410A    Fluke 289
5 Volts    5.0002    5.0006
 1 mA     0.99994     1.003
999.4?    999.23k    999.3k
9.996k?    9.9961k    9.999k
99.97k?    99.976k    99.99k

A few days ago I was contacted by Doug Malone of VoltageStandard.com, he noticed my post   and felt that the Penta-Ref should not experience a problem with the battery levels that I listed.  He requested that I return it for adjustment/repair.  So, I’ve sent it, and the DMM-Check to Doug for calibration.  I’ll post new results when I get the units back from Doug.

That’s great customer service!

Also, it looks like there’s a new model coming, the DMM-Check Plus.  It will include a 5VAC rms voltage reference,  1mA rms AC current reference, 100Hz precision frequency source, and 0.1%, 10ppm 100OHM precision resistor

Note that on the Geller SVR, he will ask you what temperature to use for the calibration.  So, that might help with your warm lab.

[T4P]

I received a pentaref some weeks ago. Me likes! But the 6 PPM/C is a little too much as my room temp seldom go under 25C. The interesting thing is that if I take the 6 PPM/C into account, the pentaref is spot on according to my calibrated Rigol DM3061 6½ digits DMM.

I'm thinking of building a "reverse oven" with a peltier to keep the pentaref at 68F (stated cal temp = 20 degrees C), but its only at the idea state. I was thinking of a peltier attached to a alu case, with the pentaref inside it, and a bigger plastic case to house everything with some thermal isolation, and then some ATTiny with some kind of temp control (maybe a good exercise in PID-control?). Ideas?

Isn't it 5½ ?

-------------------------------------------------------------------------------
Surely would work well to tune this :
http://www.kmitl.ac.th/~kswichit/DVM7135/DVM7135.htm 
Nothing beats a 8 digit led display 

No, 2 400 000 counts. You are probably thinking of DM3051 (I think it is named so) with 480 000 counts.

Edit: Your link shows a 2*4½ digit DMM, not a 8 digit one!

Oops , my bad .

Well , i did say 8 digit led display not 8 digit DVM

[SeanB]

I built a reference using an AD581, at least more accurate than the 3.5 digit meters I have. Mains powered, and in 2 boxes, one being an old wall wart providing a regulated 15V supply and the other the reference with the sockets. Built mostly from recycled parts, only thing bought was the AD581. Was based on an article in Silicon Chip, but using a different IC and a non trimmable unit. Seems stable enough for my purposes, at least at room temperature of around 25C.

[saturation]

Amplifier.cd has good documentation of his work here:

http://www.amplifier.cd/Technische_Berichte/Spannungsreferenzen/voltage_reference_build_up.html

Building a temp controlled chamber will dampen most daily variations [ shown here as the fast oscillations] but just know the slowly changing averages will plateau but still move with the seasons.  The upper and lower limits should be within spec.

At 1000 hrs.  From Maxim, demonstrates individual references will have different responses to aging, and changes in absolute voltage.  The job of the volt-nut is to find how your reference performs.




At 365+ days, see attached photo, as example of seasonal variations.

[nukie]

My 5v geller is very sensitive to input voltage changes. 1v difference to the 15v input will change the reading quite a lot.

On a similar note, the Geller SVR provided a steady 10.0000 volts output using an input voltage as low as 10.945V.

Note that on the Geller SVR, he will ask you what temperature to use for the calibration.  So, that might help with your warm lab.

[saturation]

At 6ppm/^oC, using 10V, that's 60uV/^oC.  It would only show on a 6.5 digit DMM.

Yahh, excatley what i was thinking about ! 
Peltier (Got a old coole/heater box redy with fan) and a arduino to controll it.
Then i will have good use of my scond hand Fluke 51II to monitor and calibrate the controller.

The interesting point  is, how much to +- 1-2 C actualty do for det reference in acurancy?
Normal room temp here is 22C.

[robrenz]

Make sure you put the 8846A in the chamber also or connection thermal emfs will bite you.

[alm]

This form of offset compensation is used for resistance measurements, where voltage comes from the internal current source, so you can just turn the current source off and measure the offset voltages. It's mostly found in the top end DMM's, which means 7.5-8.5 digits these days. Back when 6.5 digit was the top end (eg. Keithley 196, HP 3456A) it was found in those meters. Someone (edit: robrenz) posted a neat hack a while ago with a switch to reverse polarity to do something similar.

Not much the DMM can do with an external voltage source. Offset voltages come from dissimilar metals. It's common to use unplated copper connections for critical measurements, although you have to make sure to remove corrosion before each measurement. Most precision equipment will have copper binding posts for this reason. Keithley, Pomona and Agilent (I think) also sell low EMF leads that use special alloys to minimize EMF. Expect them to be expensive.

That said, shouldn't the two connections (plus and minus) cancel each other out in this case? Assume the two input jacks on each instrument are at the same temperature (temperature gradients should be no worse than without the oven). On each side the connections are copper (positive wire) -> tin/gold (plating) -> copper (PCB traces/internal components) -> tin/gold (plating) -> copper (negative wire), so the net offset voltage should be zero.

[robrenz]

I am not an expert so these are opinions.

Thermal Emfs are allways present, it is just a matter of whether they are a significant portion of your measurement value.  That is why they become important at low resistance measurements because you can be measuring at micro or nanovolt levels.  If you a measurement entails single digit micro Volt levels, Thermal emfs can be larger than the voltage you are trying to measure. Only current measurements will be somewhat immune because the meter is measuring its internal shunt voltage and the thermal EMF contributions of that circuit should be accounted for in the meter design and in the calibration. (Shunt temperature rise will still cause thermal emf contributions). Resistance and voltage are directly affected and 4 wire ohms does not get rid of it.  Resistance is a voltage measurement and offset ohms or true ohms techniques eliminate the thermal emf errors of resistance measurements.

IMO Since the thermal EMF values vary significantly with levels of oxides present and many plating's use a sub plating that may not be mentioned it is shaky ground to assume what they actually are and that they will cancel.

I posted the current reversal setup alm mentioned but I since realized I can do offset compensated ohms on the 8846A also. I checked with Fluke and my procedure is valid. I intend to do a post on the technique soon.

[alm]

My initial impression is that EMF should not be worse than without the oven, but I would verify this assumption before relying on it. The kind of plating shouldn't matter, as long as the two leads are identical. Difference in metal composition may be an issue, however. Maybe borrow some 'oxygen-free' copper cables from a budding audiophile? Shorting the terminals within the oven should give you an estimate of the offset voltages.

The LTZ1000 datasheet only mentions that temperature gradient across the traces to the voltage reference should be minimized, and that the traces should have the same dimensions so heat loss is about the same, not that gradients between the reference and the PCB should be minimized (hard to do, since the reference is kept at >= 60°C). Granted, the number of metal-metal transitions for a refence on a PCB is much smaller than between two pieces of equipment.

I posted the current reversal setup alm mentioned but I since realized I can do offset compensated ohms on the 8846A also. I checked with Fluke and my procedure is valid. I intend to do a post on the technique soon.

I edited my post to properly credit you.

[SeanB]

If you want to see thermal effects easily, take any analogue ammeter ( 1mA to 500mA) and heat one of the terminals a little with a soldering iron. You will see a deflection on the meter.

[billclay]

On a similar note, the Geller SVR provided a steady 10.0000 volts output using an input voltage as low as 10.945V.
Geller SVR 23.5° C
Value    34410A    Fluke 289
10 Volts    10.0000    10.000

Did you not test the SVR using a 6 1/2 digit readout on the 34410A?

[Circuitous]

Did you not test the SVR using a 6 1/2 digit readout on the 34410A?

I originally didn't have it setup for full resolution.   A few days after posting the tests, I was contacted by Doug Malone of VoltageStandard.com, he noticed this post and felt that the Penta-Ref should not experience a problem with the battery levels that I listed.  He requested that I return it for free adjustment/repair.   

I mailed the units back USPS on the evening of 4/3, and received the recalibrated units on Monday 4/9, that was fast!  I let them sit on the bench overnight to acclimatize before testing.

 I re-ran the tests , I've posted most of the info below.
To test the drop-out voltage on the PentaRef, I temporarily replaced the batteries with 2 isolated bench power supplies.  The PentaRef provided stable accurate output, identical to the battery readings (below) when run at voltages from 9.5V down to 6.5V.  It did not matter if one supply was set at 6.5V and the other at 9.5V, the reference output remained the same.

Test setup:

• All tests were performed at  21° C.
• The Agilent 34410A was set to manual range for each test, with NPLC set to 10, and nulled.
• The Fluke 289 was also set to manual ranging and nulled for each test.
• One pair of Agilent probes was used for all tests, and was moved between meters for each set of tests.
• The PentaRef and DMM-Check units were powered by fairly new 9V batteries, providing 9.2V under load.
• The Geller SVR was powered by a Rigol 1308A set at 15.00 Volts.

Geller SVR
Value    34410A    Fluke 289
10 Volts    10.00005    10.001    

Penta-Ref
Setting    34410A    Fluke 289    
0.2500    0.250036    0.2500    
0.4900    0.489981    0.4900    
0.5100    0.509995    .5100    
4.9000    4.89989    4.9001    
10.0000    9.99986     10.000    

DMM-Check
Value    34410A    Fluke 289    
5 Volts    4.99994     5.0001
1 mA    0.999923    1.000
999.2?    999.231?    999.2?    
9.996k?    9.99605k    9.996k    
99.97k?    99.9746k    99.97k

I may re-run the tests again in a month or two to check for drift.

Overall, these all good products and both Doug Malone and Joe Geller provide great support.

[T4P]

Looks like your 289 is a smidgen less accurate but it perfectly falls in spec since the 34401A is so bloody stable .

[Joe Geller]

Hi,  Just checking in, I heard there were some questions about our GELLER SVR product.

nukie: >>My 5v geller is very sensitive to input voltage changes. 1v difference to the 15v input will change the reading quite a lot.

  Our SVR boards are just very high quality platforms for the AD586/AD587 reference chips.  The data sheet for the AD586 (5V) says the limit for line regulation is +/- 150 uV / V, most do much better.  That is why we ask users to set 15.0 V for calibration activities.  Most hobby and amateur science labs can set 15.0 V to within 100 mV or better (3 1/2 digits).  Please remember, the whole idea of the SVR program is very low cost voltage transfer at a single voltage point.  All of the caveats, setting 15.0 V, using it at the temperature where we do the calibration (user specified, now that we have the micro-environmental chamber, warm up time, keeping away from a transmitting cell phone, out of direct sunlight (for thermal gradiants) are just to help you get as close as possible to our calibration point.

  That said, if you pay the actual return postage, I would be happy to have a look at your 5V SVR board (shipping address at our website, or send an email to me joegeller -at- gellerlabs.com.

Circuitous: 34410A 10.000 05 V, same offer.  Normally we charge $10 plus return shipping.  However, if you will pay actual return shipping, send your 10 V SVR back to us and I will run it through the micro-environmental chamber with the Null meter and see how it is doing (For the 10 V boards, we now also report the actual measured temperature coefficient for +/- 1c, it takes several hours to measure since the small chamber test box and SVR board need to heat soak to the new temperature for a good measurement at the ppm level.)

  I think there is a lot of space in this market for a range of products.  We believe our niche is at single voltage points and getting as close as possible (and trying to do it really really well) with a product priced for the amateur scientist / electronics hobbyist.

  Some of the parameters to look at when selecting a voltage reference chip / product are:

change in output reference voltage with:

input voltage change
output current loading
time
temperature
humidity (usually a non-issue, even with the plastic packages for > 1 or 2 ppm)
and output noise <0.1 Hz ("DC" like changes and shifts) >0.1 Hz to 100 Hz and >100 Hz

I'm probably forgetting one or two, but those are the sorts of parameters we consider in our product designs.  In this price class, once a near optimal chip is found, the answer is probably not to get all of those parameters to near "zero", but to deal with them, such as by defining the actual change with temperature (temperature coefficient or tempco) near where the product will be used.

[saturation]

Great to read you here, Joe.  I have one of your units now going on its 2nd year and its holding cal nicely, although I will send it in for a check later on.  I've been tracking its performance nearly continuously all that time.  Here are some added observations I hope you can check:

SVR response to barometric pressure: while I can control humidity to some extent in a small chamber, there can be ~ < 100uV variations when the pressure is falling, aka rain or snow is coming.  If the humidity is left to rise above 50% RH the variations are even wider.  Its erratic as the pressure falls, and stabilizes at higher than calibration levels once the nadir is achieved.   Its less sensitive to pressure rises above 30in Hg or 1016hPa.

tens of uV rise with ambient temperature [ consistent with data sheets, 5-20 ppm/^oC]

The bolded items below can be controlled fairly easily by most labs, but to eliminate environmental variations and given my ~ 2 year characterization of your board,  doing cal checks when ambient conditions of temp, pressure, and humidity are close [i.e, best if nearly identical] to when the your original cal was done practically eliminates the variation from these effects.  It also eliminates the need to spend resource building a chassis to control against these variations, unless one cannot wait for the right weather to do the calibration check.  Thus, the only drift remaining is that of the units age.

Some of the parameters to look at when selecting a voltage reference chip / product are:

change in output reference voltage with:

input voltage change
output current loading
time
temperature
humidity (usually a non-issue, even with the plastic packages for > 1 or 2 ppm)
and output noise <0.1 Hz ("DC" like changes and shifts) >0.1 Hz to 100 Hz and >100 Hz

I'm probably forgetting one or two, but those are the sorts of parameters we consider in our product designs.  In this price class, once a near optimal chip is found, the answer is probably not to get all of those parameters to near "zero", but to deal with them, such as by defining the actual change with temperature (temperature coefficient or tempco) near where the product will be used.

[Joe Geller]

saturation>>SVR response to barometric pressure: while I can control humidity to some extent in a small chamber, there can be ~ < 100uV variations when the pressure is falling, aka rain or snow is coming.  If the humidity is left to rise above 50% RH the variations are even wider.

I think your SVR board is very broken and/or there is a problem with the test setup.  With many hundreds of SVR boards in the field, there have been no other reports of ~10 ppm (100 uV) variations with normal swings in atmospheric pressure and humidity (short of condensation, another matter).  Max tempco for the "L" grade units is 5 ppm/c, most do better.

[saturation]

Yes, I've controlled for lead effects, thermoelectric effects, stray RF, etc.,

OK, Will send it back to you, Joe.  FWIW the identical test setup is used on a Krohn Hite MV106 and never a problem albeit the tracking is only about 5 months of data.  The MV106 is spec'd to < 5uV stability [ I get it practically to 1uV or less under controlled environmental conditions] and resolution to <= 100nV and very little variation through the winter and current summer.

There is no problem using SVR to resolutions of 10.000 VDC, but at 10.000 00 VDC controlling the uV level is subject to more variation.  But we shall see after I return it for a cal check.

saturation>>SVR response to barometric pressure: while I can control humidity to some extent in a small chamber, there can be ~ < 100uV variations when the pressure is falling, aka rain or snow is coming.  If the humidity is left to rise above 50% RH the variations are even wider.

I think your SVR board is very broken and/or there is a problem with the test setup.  With many hundreds of SVR boards in the field, there have been no other reports of ~10 ppm (100 uV) variations with normal swings in atmospheric pressure and humidity (short of condensation, another matter).  Max tempco for the "L" grade units is 5 ppm/c, most do better.