Building a 7 decade voltage calibrator

[gerrysweeney]

Does anyone know of any projects to build a high quality seven decade voltage calibrator?  I am thinking about having a go at building one but was wondering if it had been done DIY with any degree of success?

Gerry

[Andreas]

Linear Technology AN86 may be a starting point.

with best regards

Andreas

[Andreas]

Another posssibility may be this:

http://www.edn.com/design/other/4326640/DC-accurate-32-bit-DAC-achieves-32-bit-resolution

But this design leaves much room for optimization.
The given values for accuracy and linearity are very optimistic.

With best regards

Andreas

[saturation]

You could modify the basic form of any one of the series of defunct pure analog EDC calibrators.  They are 6 decades, but have great specs and haven't been made since the 1980s.  Dave reviewed and tore down the MV106 version some time back.



Here's the full manual, includes schematic and theory of operation, for one variant that goes from 300VDC to 300mVDC ~ 30ppm 1 year accuracy and 5-10ppm ?? 1 hr transfer accuracy.  Note those specs are ~ identical to all models EDC produced as voltage references.

http://ko4bb.com/Manuals/09)_Misc_Test_Equipment/Krohn-Hite_VS330_DC_Voltage_Standard_Operator_Manual-1979&Before.pdf

[JoeyP]

Also take a look at the Data Precision 8200. They've been out of production for a while, but full schematic can be found on line. They use mostly cheap simple CMOS switches to get surprisingly good results.

[quarks]

Hello Gerry,
that will be a very interesting DIY project. What are your thoughts so far about the voltage ranges, accuracy and stability you like to achive?

If you want the best possible result, you probably want to use a LTZ1000 reference.
Then have a look here
https://www.eevblog.com/forum/projects/ultra-precision-reference-ltz1000/msg180731/#msg180731
and for some ideas about Kelvin Varley Devider
look here
https://www.eevblog.com/forum/projects/kelvin-varley-divider-(and-precision-voltage-source)/msg70859/#msg70859
And look into the articles from Conrad Hoffman (also member of this forum) about metrology, KVDs, Hamon Divider and bridges.
http://conradhoffman.com/electronics01.htm
http://conradhoffman.com/mini_metro_lab.html
http://conradhoffman.com/HamonResistor.html

bye
quarks

[Andreas]

Also take a look at the Data Precision 8200.
They use mostly cheap simple CMOS switches to get surprisingly good results.

In this design the quality of the switches does nearly play no role by design.
But you will need many (>15) highest quality resistors (hand trimmed for accuracy + linearity) to achive the good results for the 10V output range.
In the mean time 20 bit DACs are available (laser trimmed) which will have same or better data.

With best regards

Andreas

[branadic]

In the mean time 20 bit DACs are available (laser trimmed) which will have same or better data.

Wow, this is pretty interesting:

DAC1282 - 24bit - http://www.ti.com/product/dac1282
DAC1280 - 24bit - http://www.ti.com/product/dac1280
DAC1220 - 20bit - http://www.ti.com/product/dac1220

AD5791 - 20bit - http://www.analog.com/en/digital-to-analog-converters/da-converters/ad5791/products/product.html

I wonder why I couldn't find parts from LinearTechnology.

[sync]

Building a n-decade voltage calibrator is on my to-do list since a long time. I think 7 decades are out of my skills and needs. I want usable 5-6 decades. Which should be hard enough. I want to experiment with amspires sigma-delta PWM¹ and Linear Technology AN86 ideas. One goal is not using resistors for anything affecting the output voltage (at the basic range). Ultra precision resistors (hybrids) are hard to get and expensive.

¹) https://www.eevblog.com/forum/projects/general-purpose-power-supply-design-7488/msg99807/#msg99807

[sync]

DAC1282 - 24bit - http://www.ti.com/product/dac1282

It has a differential voltage output. So an ultra precision differential amp is needed, right?

DAC1280 - 24bit - http://www.ti.com/product/dac1280

Differential current output. Worse than above.

DAC1220 - 20bit - http://www.ti.com/product/dac1220

0-5V single ended output - fine.

AD5791 - 20bit - http://www.analog.com/en/digital-to-analog-converters/da-converters/ad5791/products/product.html

A "high" voltage DAC. It can use a 7V Reference directly. Nice!
Edit: Wow! These are really expensive (60$ or 100$).

[Andreas]

0-5V single ended output - fine.

AD5791 - 20bit - http://www.analog.com/en/digital-to-analog-converters/da-converters/ad5791/products/product.html

A "high" voltage DAC. It can use a 7V Reference directly. Nice!
Edit: Wow! These are really expensive (60$ or 100$).

The AD5791 is the one that I had in mind.

Compared to the around 11 high precision resistors (10-60$ each depending on long term stability) that it replaces + the trimming time + the equipment for trimming the 100$ are a good invest.
All resistors on the single chip have same temperature and this results in < 0.05 ppm/°C temperature drift which cannot be done with discrete resistors without a temperature stabilized oven.
And the 1ppm linearity spec is better than many 6,5 digit precision instruments have.

With best regards

Andreas

[sync]

It's surely a very nice IC. But I'm not comparing it with a discrete resistor solution. That's just impractically. I hope that an AN86 type solution will give similar results.

I would using the AD5791 directly with a 7V reference. The resolution would be 6.7uV. For the calibrator I would use 10uV resolution. Actually i find the 6.7uV too coarse for this. The AD5791 has 1ppm typ. and 3ppm max. INL. That is an error of 0.7 typ. and 2.1 max. "counts".

Unfortunately the LTC2400 can not used with a 7V reference. I would use a capacitive divider to get 3.5V. It has 4ppm typ. and 15ppm max. INL. With the same 10uV calibrator resolution it's an error of 1.4 typ. and 5.25 max. "counts".

The LTC2400 has also the advantage of a much finer resolution. And the AD5791 has a DNL up to 2.5 LSB. If i understand this correctly it's not guaranteed to be monotonic.

Please correct me if I'm wrong or wrote complete nonsense.

[saturation]

Just a side comment, if you consider building around the EDC design, you have a proven performer tested in the field by industrial users for several decades because said models contributed to the economic viability of the EDC Corporation for many decades plus allowed them to create and sell variations of these voltage calibrator models based on adjusting the basic modules.  Thus, the design is fairly flexible and proven.  The components are fairly common, but they will need replacement with modern versions. 

[branadic]

As far as I can see on the picture of the evalboard:



they have spent a voltage reference circuit based on the LTZ1000.  What is missing is the schematic for that specific part of the circuit. I'm sure that they didn't use the

Recommended External Reference for the AD5791

• For a precision, ultralow noise, 5V reference, we recommend the ADR445 or the ADR4550.

for characterization.
Up to now I couldn't find a diy project with LM399/LTZ1000 and the AD5791. Could be worth developing such a circuit.

[mrflibble]

Regarding the AD5791, this related appnote might be of interest as well.

[quantumvolt]

Just a side comment, if you consider building around the EDC design, you have a proven performer tested in the field by industrial users for several decades because said models contributed to the economic viability of the EDC Corporation for many decades plus allowed them to create and sell variations of these voltage calibrator models based on adjusting the basic modules.  Thus, the design is fairly flexible and proven.  The components are fairly common, but they will need replacement with modern versions.

This is correct. However, all (?) these EDC/Krohn-Hite boxes are nothing but a super low tempco zener diode and a chopper amp. And add a fortune in precision resistors. Building the reference is cheap (1N829A and LT1050), building the multi decade rotary switch resistor divider is hundreds of dollars. The schematics of two different versions (positive / negative supply for the current source for the ref diode) is in a tear down video and on kobb - both linked on this forum.

[Andreas]

I would using the AD5791 directly with a 7V reference.

Unfortunately the LTC2400 can not used with a 7V reference. I would use a capacitive divider to get 3.5V. It has 4ppm typ. and 15ppm max. INL. With the same 10uV calibrator resolution it's an error of 1.4 typ. and 5.25 max. "counts".

The LTC2400 has also the advantage of a much finer resolution.

Please correct me if I'm wrong or wrote complete nonsense.

I would use a capacitive voltage multiplier * 1.5 having a 10.5V maximum range.
For the LTC2400 I would use a voltage divider * 0.666 having a 4.6V reference and up to 5.1V measurement range.

See:
ppmgeek

The LTC2400 linearity is given as "best straight fit" so if you do a "end point calibration" of the range the actual value is practically doubled against the datasheet value. So 4 ppm are a 40uV missing to the 2.5 V mid range value.
(You can find this in carefully analyzing the assembler code of AN86).

The LTC2400 has a noise voltage of 10uVpp (= 0.3ppm, eff in best case). So the settling time to the final value below 10uV will need averaging many measurement values. So the much finer resolution is only valuable when you can use a integration time of 1 minute or longer.

So the settling time of the AD5791 is much better. And if the switching noise between neigboured output values is low enough one could have the idea to do some PWM with 10uV neighboured values.

With best regards

Andreas

[Andreas]

However, all (?) these EDC/Krohn-Hite boxes are nothing but a super low tempco zener diode and a chopper amp. And add a fortune in precision resistors. Building the reference is cheap (1N829A and LT1050)

Thats not all. You need very specialized know how and a lot of work to build a good device with theese cheap parts.

First you have to age and select some out of a bunch of the zeners for ageing and noise. This will take around 1 year and from hundreds of zeners there will be few dozens left. Then you have to determine "zero tempco current" and build your current source accordingly for each device. And perhaps there are more treats to do that I still do not know.

With best regards

Andreas

[branadic]

I would use a capacitive voltage multiplier * 1.5 having a 10.5V maximum range.

This means the use of 2x LTC1043, one half for the divide by 2 and another complete part for the multiply by 3 section.
And what about the negative voltage reference rail? Just a simple inverter with gain=1?

Edit: The second half of the first LTC1043 could be used as an ultra precision voltage inverter.

[quantumvolt]

However, all (?) these EDC/Krohn-Hite boxes are nothing but a super low tempco zener diode and a chopper amp. And add a fortune in precision resistors. Building the reference is cheap (1N829A and LT1050)

Thats not all. You need very specialized know how and a lot of work to build a good device with theese cheap parts.

First you have to age and select some out of a bunch of the zeners for ageing and noise. This will take around 1 year and from hundreds of zeners there will be few dozens left. Then you have to determine "zero tempco current" and build your current source accordingly for each device. And perhaps there are more treats to do that I still do not know.

With best regards

Andreas

Also correct from a purist point of view. However - for those of us that can only afford to buy one diode (mine was USD10 +shipping 4 on fleabay) the knowledge that the current will be around 7.5 mA for all diodes and that the tempco is monotonic increasing in current leaves running in the device to a question of investment of free hobby time. Just set the current approximately, log the tempco behaviour and adjust the current up or down accordingly. Iterate for a year or more. Typical diagram page 2 in http://www.aeroflex.com/ams/Metelics/pdfiles/1N821-1N829-1_TC_Zener.pdf

[grego]

EDN had a video on building a 5-decade KV reference - neat as a starting point:

http://video.linear.com/28

[quantumvolt]

...

 Then you have to determine "zero tempco current" and build your current source accordingly for each device. And perhaps there are more treats to do that I still do not know.

...

Determining the current is easy due to the monotonicity in tempco vs current. Changing the current source is only adjusting a trimpot. I have seen two different versions of the schematics. No more magic. The sum of Rt and the trim determines the current, and the 'strange' bias/feedback network is the same for both schematics. Maybe there is matching of transistors, but that is not mentioned in the manual.

[Andreas]

I would use a capacitive voltage multiplier * 1.5 having a 10.5V maximum range.

This means the use of 2x LTC1043, one half for the divide by 2 and another complete part for the multiply by 3 section.
And what about the negative voltage reference rail? Just a simple inverter with gain=1?

Edit: The second half of the first LTC1043 could be used as an ultra precision voltage inverter.

Why using 2 LTCs for a factor 1,5 multiplication?
with 1 LTC1043 you can: Divide by 2 + add the divided voltage to the input voltage.

The inverted voltage for 10V will not work due to limited power supply range. (maximum 18V).
For what do you need it?

With best regards

Andreas

[EEVblog]

You could modify the basic form of any one of the series of defunct pure analog EDC calibrators.  They are 6 decades, but have great specs and haven't been made since the 1980s. 

The problem with this is that the switches are really schmick. Not sure where you'd get suitable ones for a 7 decade box. I greatly doubt youd get good 6 decade performance with run of the mill switches, let alone 7.

[EEVblog]

And perhaps there are more treats to do that I still do not know.

The switches. EDC didn't just go down to Tricky Dicks and pick those up. They could be the most expensive thing in the whole system, by far.