Re: [volt-nuts] The averaging reference
On Sat, 20 Dec 2014 22:17:57 +0100 Joel Setton set...@free.fr wrote: I wasn't aware of the degraded long-term drift performance in the plastic packages, as compared to the metal can. I'm surprised they can't protect the chip from package-induced effects! Well, that's where physics strikes back... and economics. None of the materials known to men are completely gas or water tight. And an exchange of gas/water leads to change of physical properties or even chemical reactions. AFIK the most gas-tight enclosure is a metal can (afaik only hydrogen and helium diffuse trough a steel can, surface does not readily react with most substances found in air, but needs non-metalic isolation for the wires going in/out),followed by glas packages (AFAIK water tight, but not completely gas tight. also can act as a getter material if outside surface is clean and in vacuum), followed by ceramics (little gas exchange with the inside, but porous, ie can store gases/water on the surface). Plastic packages are mainly one thing: cheap. Neither gas nor air-tight, they even store a lot of chemical compounds from production within the material, that slowly leaks out. Also they are quite hygroscopic, to the extend that chips are backed out before soldering, in order to prevent the vaporizing water from breaking the chip. Even small changes in the composition of the plastic can change the pin-to-pin resistance from 10M to 2M. For normal electronics this doesn't matter, but here... So, yes, it is possible to have better packages than just plastic. But it is not economical to keep these around for the one or two people a year who actually need them. (well, they do it with space grade components, but they charge you 1000 times the price of the commercial equivalent). Attila Kinali -- I pity people who can't find laughter or at least some bit of amusement in the little doings of the day. I believe I could find something ridiculous even in the saddest moment, if necessary. It has nothing to do with being superficial. It's a matter of joy in life. -- Sophie Scholl ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.
Re: [volt-nuts] The averaging reference
The Datron 1081/1082 are 7-1/2 digit meters. The zeners are two pair of two zeners in series to get the ref voltage up high enough. In the two 1081s I had, they worked quite well, ref’d to the two Fluke 732As I had at the time. But not 8-1/2 digit grade, I don’t think. I have no idea what the 127x/128x sereis use; probably the LTZ1000A... ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.
Re: [volt-nuts] The averaging reference
Jan, Thanks for a good summary f the pros/cons. Of course the LTZ1000 is much closer to the current state of the art, but the REF102 is far easier to use and to calibrate. I'm definitely not shooting for sub-ppm performance, if I can build anything that stays within (say) 20 ppm long-term, that would be more than adequate as a home standard. I wasn't aware of the degraded long-term drift performance in the plastic packages, as compared to the metal can. I'm surprised they can't protect the chip from package-induced effects! One thing I don't like about the LM199 and LTZ1000 is that although they are stable, they are sold uncalibrated. As a result, building a 10-V reference with either of them would require at least two very stable resistors, one of which must be selected within a range of several percent to get an accurate 10V output. Most of the DVMs I have seen with the LM199 / LTZ1000 use soft calibration in which the calibration coefficient is stored in memory, and the voltage measurement is performed in ratiometric mode. Building a 10V reference is a rather different problem. As before, comments and suggestions will be welcomed! Joel Setton On 19/12/2014 19:28, Jan Fredriksson wrote: It's no coincidence that virtually all 8.5 digit DMMs use the LTZ1000. It's in a class of it's own. REF102 is not in the same class, even if you average a handful. But there are a couple of nice things about the REF102, though for more moderate requirements - You get a reference at 10V, +/-0.0025V, trimmable (not a 5% 7V of the LTZ1000) - Moderate power / current - Low sensitive to supply voltage - Very simple to implement There was a metal can version but it's obsolete. But be aware that the TI site still shows the metal can spec 5ppm/1000h while the available packages are actually 20ppm/1000h! ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there. ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.
Re: [volt-nuts] The averaging reference
I believe the 1281 uses the LTZ1000. The datasheet mentions dual references. My 4920 has both a ltz1000 and a lm399. I think you can find pics of the 1281 dual ref setup on eevblog. I would not be surprised if the 8508a has a similar configuration. Todd On Saturday, December 20, 2014, Richard Moore richiem5...@gmail.com wrote: The Datron 1081/1082 are 7-1/2 digit meters. The zeners are two pair of two zeners in series to get the ref voltage up high enough. In the two 1081s I had, they worked quite well, ref’d to the two Fluke 732As I had at the time. But not 8-1/2 digit grade, I don’t think. I have no idea what the 127x/128x sereis use; probably the LTZ1000A... ___ volt-nuts mailing list -- volt-nuts@febo.com javascript:; To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there. ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.
Re: [volt-nuts] The averaging reference
On 20 Dec 2014 21:18, Joel Setton set...@free.fr wrote: Jan, Thanks for a good summary f the pros/cons. Of course the LTZ1000 is much closer to the current state of the art, but the REF102 is far easier to use and to calibrate. I'm definitely not shooting for sub-ppm performance, if I can build anything that stays within (say) 20 ppm long-term, that would be more than adequate as a home standard. One thing I don't like about the LM199 and LTZ1000 is that although they are stable, they are sold uncalibrated. As a result, building a 10-V reference with either of them would require at least two very stable resistors, one of which must be selected within a range of several percent to get an accurate 10V output. I suspect if you built something very stable using an LTZ1000, it would be possible to get one or more volt-nut with a 3458A or similar to measure it for you. You could even average the result from several volt nuts. Dave. ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.
Re: [volt-nuts] The averaging reference
___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.
Re: [volt-nuts] The averaging reference
Guys, Thanks for your comments! The idea is to use a simple resistor averaging network, as shown in Fig. 14 of the REF102 datasheet. I hadn't thought of the effect of relative humidity, would the temperature-controlled enclosure at (say) 50C/122F change things in this respect? One nice thing about this circuit topology is that it's a simple matter to measure the drift of each individual reference chip with respect to the averaged output, even plot it over temperature. At this time my biggest question is how to compute the total error budget for such an arrangement. Any help would be appreciated! Joel Setton ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.
Re: [volt-nuts] The averaging reference
It's no coincidence that virtually all 8.5 digit DMMs use the LTZ1000. It's in a class of it's own. REF102 is not in the same class, even if you average a handful. But there are a couple of nice things about the REF102, though for more moderate requirements - You get a reference at 10V, +/-0.0025V, trimmable (not a 5% 7V of the LTZ1000) - Moderate power / current - Low sensitive to supply voltage - Very simple to implement There was a metal can version but it's obsolete. But be aware that the TI site still shows the metal can spec 5ppm/1000h while the available packages are actually 20ppm/1000h! ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.
Re: [volt-nuts] The averaging reference
On 19 Dec 2014 19:30, Jan Fredriksson j...@41hz.com wrote: It's no coincidence that virtually all 8.5 digit DMMs use the LTZ1000. It's in a class of it's own. What do the 8.5 digit meters use if they don't use the LTZ1000? Dave. ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.
[volt-nuts] The averaging reference
Folks, In my search for the Perfect Volt, I'm thinking about building a reference voltage generator which would average the voltages generated by 8 or 10 REF102CP chips (with a simple resistor network), mounted in a temperature-controlled box. Among all the experts on this list, does anyone have any experience with such a project? It looks good on paper, but what kind of stability can I reasonably expect? Many thanks! Joel Setton ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.
Re: [volt-nuts] The averaging reference
Hello, I would expect a stability of around 10-15 ppm over one year after 6-12 months run in time. The reason is that with the plastic package all references will have about the same change in the order of 0.5ppm/% rH change. So assumption is that you have maximum 30% rH change over one year. With best regards Andreas Am 18.12.2014 um 16:41 schrieb Joel Setton: Folks, In my search for the Perfect Volt, I'm thinking about building a reference voltage generator which would average the voltages generated by 8 or 10 REF102CP chips (with a simple resistor network), mounted in a temperature-controlled box. Among all the experts on this list, does anyone have any experience with such a project? It looks good on paper, but what kind of stability can I reasonably expect? Many thanks! Joel Setton ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there. ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.
Re: [volt-nuts] The averaging reference
In message 5492f59e.8060...@free.fr, Joel Setton writes: In my search for the Perfect Volt, I'm thinking about building a reference voltage generator which would average the voltages generated by 8 or 10 REF102CP chips (with a simple resistor network), mounted in a temperature-controlled box. The problem with averaging a bunch of the same type of chip is that you cannot average out the systematic errors of that chip. If you want to do it right, you should average all sorts of different (but obviously: good) chips, possibly assigning them different weight depending on their qualities. -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 p...@freebsd.org | TCP/IP since RFC 956 FreeBSD committer | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence. ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.
Re: [volt-nuts] The averaging reference
On 18 Dec 2014 19:06, Poul-Henning Kamp p...@phk.freebsd.dk wrote: In message 5492f59e.8060...@free.fr, Joel Setton writes: In my search for the Perfect Volt, I'm thinking about building a reference voltage generator which would average the voltages generated If you want to do it right, you should average all sorts of different (but obviously: good) chips, possibly assigning them different weight depending on their qualities. I was aware of three 8.5 digit multimeters * Keysight 3458A * Fluke 8508A * Keithley 2002. But http://www.eevblog.com/forum/testgear/8-5-digit-dmm/ lists several others - some from manufacturers I have never heard of. I would assume that they all use a state of the art voltage references. I wonder if they all use the same method? If not, perhaps averaging those would be one approach. How would you average them? Assuming they are supposed to be the same voltage, would it be OK to tie N output together with N resistors? Use low value resistors on the references you trust most, and higher values on those you trust least. Dave ___ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.