Question: Open Hardware Multislope. Any out there?

[Pajeh]

Hello everybody!
I'm a big fan of Dave's videos and was thinking about his teardown of the Agilent TrueVolt 34461A 6.5digit DMM:
https://www.eevblog.com/forum/blog/eevblog-485-agilent-truevolt-34461a-multimeter-teardown/

Particularly about the multislope technique. I looked up the differences in Delta-Sigma and the multislope for the high end equipment.
I understand the basic principle of the D-S and the Multislope and read that the linearity of the D-S isn't as good. People also say that for the mid to lower end the multi-slope should be even less expensive, but yeah... not sure about that. (D-S get cheap but errors and uncertainty discussion can be huge here)
Can anyone point me in the right direction?

Which leads to me wanting to try it out!
My reference here is of the HP 3458A. Don't know which version of the multislope it is, but I downloaded the HP 3458A CLIP and read some about it. The patents of the multislopes up to the III (IV is still protected) aren't valid anymore, so no problems in OpenHardware / other licenses. Of course they implemented it well and are a reference, but people copy! How good do these copies turn up to be?
Again I find some information about it, but no open source stuff with hard measurements. I went through some OpenHardware DMMs which talk about it, but didn't find circuit discussion and certainly no good test.
Is there any multislope implementation (with tests) out there?

If yes, I would like to read/learn more about it. If not, why isn't there? Isn't it worth it?
The simpler way to look at it would be:
hey, sample 3 op amps, buy some weird resistores, get your favorite microcontroller, FPGA or alike plus some 20 MHz time measurement part, buy an LTZ1000 replacement board and build a 8.5 Digit Measurement setup (not saying DMM to leave all protection switching and so on aside)

Thanks for inputs in advance!

[Berni]

Last time i needed precision measurement i used a ADS1282. Worked pretty well and its a single chip solution pretty much. But on the down side it is a very expensive chip.

[Pajeh]

Last time i needed precision measurement i used a ADS1282. Worked pretty well and its a single chip solution pretty much. But on the down side it is a very expensive chip.

Thx for the tip. Looks like a good chip. The INL is down to 4 ppm, BUT I'm not trying to find an specific chip for a tasks. I'm just learning the high end ADC to understand it better. Which one achieves better uncertainties nowadays and why?

[Berni]

Yeah that's different if you want to learn from it. I needed to make a product work so i didn't have time to muck about with something like that.

Looks like a fairly doable thing tho since from the looks of modern bench DMMs there are no special fancy components needed to build one. Might be a good idea to look at patents and research papers on the details of how the most accurate ones are made.

[Pajeh]

Yeah that's different if you want to learn from it. I needed to make a product work so i didn't have time to muck about with something like that.

Looks like a fairly doable thing tho since from the looks of modern bench DMMs there are no special fancy components needed to build one. Might be a good idea to look at patents and research papers on the details of how the most accurate ones are made.

=)
Yep! It looks doable, that's why I was asking myself "why don't I find any builds?"
The lack of work in the area is still a question for me.. I'll also ask around in my university.


I will probably dive into the HP 3458A CLIP and understand the schematics. There are lots of practical implementation details that I can't grasp (talk about dimension it).
I'm interested in building one to check linearity, uncertainties, etc.

First steps would be to look for papers. The patents I have checked just the US, but this one is already newer than the HP 3458A:
http://www.google.de/patents/US5148171

Think it should be lapsed worldwide.

[unitedatoms]

If you will not find a multislope as recent Open Source, look at classic old schematics. Most or all old metrology devices had their schematics opened by default.

I wonder myslef why there is no Open Source for it. People are making multi-inch telescopes by bare hands, with dozen nanometers accuracy of the mirror. This is about 8.5 digits of machnical dimenstion to tolerance ratio.

Cost contrasts: LTZ part costs x100 times less than finished 8.5 digit device.
Design basics: Seemingly simple.
A person with access to internet can implement such and such modules, software, aspects ...?: Yes
Profit!

I think such project is very neccessary for community. I envision use of unexpected materials, parts and solutions. Like polished granite base, varieties of ovens, barometrically isolated chambers, machined aluminum subassembly boxes, handwoven resistors, miniboards with network of handpicked compensating thermistors, interesting low noise power sources etc. There is the world off unknown dielectric absorption properties of contemporary plastics which may exceed Teflon (I hope), and CADs with home made laser cutters to make capacitors from scratch. Also stock of Teflon caps is very old now, may be the properties of Teflon keep improving with age.

I believe I saw the Russian KT101KT1A chip used as a reference in dual slope. The point was that inverted emitter works like a zener and there is collocated identical transistor on same die which can work as a thermal sensor.

My personal idea about reference was DiFet OpAmp with CR2032 or some carefully selected button battery in chamber. The battery can stay alive for many years with projected drift. And reference can be in a solid walled hermetic chamber.

[fcb]

An 8.5 digit OS DMM is certainly a possibility - but how on earth would you be able to calibrate it to level of accuracy/certainty that makes sense?

Keysight still make and sell the 3458A (I got a quote last week of one for a project I'm doing £6183+VAT!), if you can make an OS equivalent AND adjust/calibrate it AND prove it for <£1K i'm interested!

[unitedatoms]

The industrial dual slopes have many features unnecessary for home lab use: GPIB - not terribly necessary, high speed modes with 100KSamples at 16bit accuracy can be skipped, 50 vs 60Hz automatic switching - no need, user knows what line frequency is in the country, many other options can be hardwired, the temperature range can be narrowed, the speed in general is not an issue for patient end user, etc.

Many of advantages of starting historically later include breakthoughs like of 0.05ppm matched metal foil dividers which appeared after HP was designed. Things may be simplified just because of this.

For many ratiometric uses like INL characterization of DACs/ADCs, there is no need to have absolute voltage accuracy, some users can go by with just 8.5 digit of resolution/noise/confidence and 5.5 digit of absolute accuracy.

For differential observations / comparative measurements some can live with just high repeatability as long as reference stays uncalibrated but has good predictable drift.

Some late historical cost breaktrhoughs can help too: power LEDs for remote heating through optical fiber, cheap supercapacitors for supply. Another historical advantage is just knowledge of aging behavior of all or some modules/parts. Many design decision could be proved to be unnecessary, so the design can be simplified.

Patents and copyright limitations: there is less limitations for individual makers as long as they don't sell builds as products, this can open some possibilities, possibly skipped at a time of HP design.

Loads of parts and experience is accumulated by water cooled PC builders. The whole thing can be built to stay in steered oil tank.