Want a stable 10V ref? Best thing to do is get a fluke 732A/B and have it calibrated regularly for a few years. BUT For the Extreme (or cheap) Volt Nut, There is another choice. Attached is the circuit of a 0.1 PPM, 10 volt transfer standard. It has ONLY ONE critical part that limits its accuracy and stability. Everything else can be made to have so little effect as to be insignificant or else can be trimmed and calibrated out without using a precision voltage source using a method similar to what the daily auto cal in the HP3458 does. It measures it's own internal 6+V reference. Every year or more, best to record the value of the nominal 6.2 volts, to measurement the ageing rate, by using a known external 10 volt standard. To do that, an accurate external way is needed to find the exact ratio of the 6.2 volts to the external 10 volt standard. This can be done by using a high resolution Linear DMM (aka HP3458A) OR a Kelvin-Varley divider (aka fluke 720 ) and null meter (aka Fluke 845). Lets assume if one wants to do 0.1 PPM things, then they have the equipment and skill needed to measure a 10v to 6v ratio to 0.1 PPM. After that ratio is known and recorded the units internal 10 volts can then be set to that same ratio without further need of the external 10V standard. The main advantage of all this extra trouble is low cost accuracy and stability. There are no critical or expensive resistor needed or anything else critical, because all the TC and long term drift are canceled out whenever a manual trim is done using an external ratio device to check its own 10 V against it's own 6.2 volts. Nothing very special is needed to build this beside care and time. To get the best performance, It does take a lot of Nut-time to match and select things. Total parts cost with a well supplied analog junk box (or EBay) can be under $10. Any good 6V type reference device could be used. I'm using a well aged and selected 1n825 because I have a lot of history with them and they work good without an oven. I have not done any mailing test on these parts but they can be Zero TCed so that they do not change 0.1PPM over normal room temperature and they show no change when powered down for short periods of time and tend to return to their original value when hit with a heat gun and cold stray, so it makes a good part to start experimenting with. ws

                      Name: ws10Vref-9-3LR.gif
ws10Vref-9-3LR.gif    Type: GIF Image (image/gif)
                  Encoding: base64

Warren, You need to do your schematics a little bit better. Besides, I think you attached the wrong one. There are a number of errors on the attached schematic. Bill....WB6BNQ WarrenS wrote: > Want a stable 10V ref? > Best thing to do is get a fluke 732A/B and have it calibrated regularly for a few years. > BUT > For the Extreme (or cheap) Volt Nut, There is another choice. > Attached is the circuit of a 0.1 PPM, 10 volt transfer standard. > > It has ONLY ONE critical part that limits its accuracy and stability. > Everything else can be made to have so little effect as to be insignificant or else can be trimmed and calibrated out without using a precision voltage source using a method similar to what the daily auto cal in the HP3458 does. It measures it's own internal 6+V reference. > > Every year or more, best to record the value of the nominal 6.2 volts, to measurement the ageing rate, by using a known external 10 volt standard. > To do that, an accurate external way is needed to find the exact ratio of the 6.2 volts to the external 10 volt standard. > This can be done by using a high resolution Linear DMM (aka HP3458A) OR a Kelvin-Varley divider (aka fluke 720 ) and null meter (aka Fluke 845). > Lets assume if one wants to do 0.1 PPM things, then they have the equipment and skill needed to measure a 10v to 6v ratio to 0.1 PPM. > After that ratio is known and recorded the units internal 10 volts can then be set to that same ratio without further need of the external 10V standard. > > The main advantage of all this extra trouble is low cost accuracy and stability. > There are no critical or expensive resistor needed or anything else critical, because all the TC and long term drift are canceled out whenever a manual trim is done using an external ratio device to check its own 10 V against it's own 6.2 volts. > > Nothing very special is needed to build this beside care and time. > To get the best performance, It does take a lot of Nut-time to match and select things. > Total parts cost with a well supplied analog junk box (or EBay) can be under $10. > > Any good 6V type reference device could be used. > I'm using a well aged and selected 1n825 because I have a lot of history with them and they work good without an oven. > I have not done any mailing test on these parts but they can be Zero TCed so that they do not change 0.1PPM over normal room temperature > and they show no change when powered down for short periods of time and tend to return to their original value when hit with a heat gun and cold stray, so it makes a good part to start experimenting with. > > ws > > ------------------------------------------------------------------------ > Name: ws10Vref-9-3LR.gif > ws10Vref-9-3LR.gif Type: GIF Image (image/gif) > Encoding: base64 > > ------------------------------------------------------------------------ > _______________________________________________ > 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.

Hi Bill, Warren, Whats the matter with it? The schematic is clear enough to me, interesting and seems quite clever to this wannabe nut. If I understand correctly, the opamp adjusts the divided-down output via the regulator so that it matches the reference. Everything is powered from the 10V output, eliminating problems with variable input voltages and opamp PSRR. I would be interested to know: - Any suggestions for the wirewound resistors? Vishay bulk metal foil resistors are not cheap but say 5ppm wirewounds are not pennies either, I don't know how many you would need before a sufficient number could be matched. - I am unclear as to the precise function of the 160k and 40M resistors. I suppose one of them shifts the output, and the other compensates the zener so it still runs at the same minimum-tempco current? The zeners do look interesting as references, I admit I have been fixated on LTZ1000s. Farnell/Newark don't stock the 1N825 but they do stock the 1N825A. <http://uk.farnell.com/american-power-devices/1n825a/diode-zener-6-5v-0-25w-do35/dp/1651077?Ntt=1N825> According to its "datasheet" <http://www.farnell.com/datasheets/332215.pdf> there are also some similar looking parts that may have have improved tempco, for example 1N829A at 5ppm/K vs 1N825 at 20ppm/K. These are also available. There are even some 2ppm/K parts but these seem unobtainium. Are there any better datasheets? For example how does the tempco itself vary with temperature? The company does not seem to have a web presence. There Regards, John WB6BNQ <wb6bnq@cox.net> writes: > Warren, > > You need to do your schematics a little bit better. Besides, I think you attached the wrong one. There are a number of errors on the attached schematic. > > Bill....WB6BNQ > > WarrenS wrote: > >> Want a stable 10V ref? >> Best thing to do is get a fluke 732A/B and have it calibrated regularly for a few years. >> BUT >> For the Extreme (or cheap) Volt Nut, There is another choice. >> Attached is the circuit of a 0.1 PPM, 10 volt transfer standard. >> >> It has ONLY ONE critical part that limits its accuracy and stability. >> Everything else can be made to have so little effect as to be insignificant or else can be trimmed and calibrated out without using a precision voltage source using a method similar to what the daily auto cal in the HP3458 does. It measures it's own internal 6+V reference. >> >> Every year or more, best to record the value of the nominal 6.2 volts, to measurement the ageing rate, by using a known external 10 volt standard. >> To do that, an accurate external way is needed to find the exact ratio of the 6.2 volts to the external 10 volt standard. >> This can be done by using a high resolution Linear DMM (aka HP3458A) OR a Kelvin-Varley divider (aka fluke 720 ) and null meter (aka Fluke 845). >> Lets assume if one wants to do 0.1 PPM things, then they have the equipment and skill needed to measure a 10v to 6v ratio to 0.1 PPM. >> After that ratio is known and recorded the units internal 10 volts can then be set to that same ratio without further need of the external 10V standard. >> >> The main advantage of all this extra trouble is low cost accuracy and stability. >> There are no critical or expensive resistor needed or anything else critical, because all the TC and long term drift are canceled out whenever a manual trim is done using an external ratio device to check its own 10 V against it's own 6.2 volts. >> >> Nothing very special is needed to build this beside care and time. >> To get the best performance, It does take a lot of Nut-time to match and select things. >> Total parts cost with a well supplied analog junk box (or EBay) can be under $10. >> >> Any good 6V type reference device could be used. >> I'm using a well aged and selected 1n825 because I have a lot of history with them and they work good without an oven. >> I have not done any mailing test on these parts but they can be Zero TCed so that they do not change 0.1PPM over normal room temperature >> and they show no change when powered down for short periods of time and tend to return to their original value when hit with a heat gun and cold stray, so it makes a good part to start experimenting with. -- John Devereux

Hi, The schematic has the opamp+ connected to the wrong point and there are some other details wrong. In another job there was an instrument that used an 1n829a run at about 7.5mA for a 10 V reference that used to hold to better than 1uV over lab temperatures. It was done using the Op amp to generate the 10V that directly drove the 7.5mA into the zener, select on test resistors. but for 0.1ppm you need to address opamp heating and unintended thermocouples everywhere. > From: john@devereux.me.uk > To: volt-nuts@febo.com > Date: Sun, 4 Sep 2011 15:56:00 +0100 > Subject: Re: [volt-nuts] Sub PPM 10V transfer standard > > > Hi Bill, Warren, > > Whats the matter with it? The schematic is clear enough to me, > interesting and seems quite clever to this wannabe nut. > > If I understand correctly, the opamp adjusts the divided-down output via > the regulator so that it matches the reference. Everything is powered > from the 10V output, eliminating problems with variable input voltages > and opamp PSRR. > > I would be interested to know: > > - Any suggestions for the wirewound resistors? Vishay bulk metal foil > resistors are not cheap but say 5ppm wirewounds are not pennies > either, I don't know how many you would need before a sufficient > number could be matched. > > - I am unclear as to the precise function of the 160k and 40M > resistors. I suppose one of them shifts the output, and the other > compensates the zener so it still runs at the same minimum-tempco > current? > > The zeners do look interesting as references, I admit I have been > fixated on LTZ1000s. > > Farnell/Newark don't stock the 1N825 but they do stock the > 1N825A. > > <http://uk.farnell.com/american-power-devices/1n825a/diode-zener-6-5v-0-25w-do35/dp/1651077?Ntt=1N825> > > According to its "datasheet" > <http://www.farnell.com/datasheets/332215.pdf> there are also some > similar looking parts that may have have improved tempco, for example > 1N829A at 5ppm/K vs 1N825 at 20ppm/K. These are also available. There > are even some 2ppm/K parts but these seem unobtainium. > > Are there any better datasheets? For example how does the tempco itself > vary with temperature? The company does not seem to have a web presence. > > There > > Regards, > > John > > > WB6BNQ <wb6bnq@cox.net> writes: > > > Warren, > > > > You need to do your schematics a little bit better. Besides, I think you attached the wrong one. There are a number of errors on the attached schematic. > > > > Bill....WB6BNQ > > > > WarrenS wrote: > > > >> Want a stable 10V ref? > >> Best thing to do is get a fluke 732A/B and have it calibrated regularly for a few years. > >> BUT > >> For the Extreme (or cheap) Volt Nut, There is another choice. > >> Attached is the circuit of a 0.1 PPM, 10 volt transfer standard. > >> > >> It has ONLY ONE critical part that limits its accuracy and stability. > >> Everything else can be made to have so little effect as to be insignificant or else can be trimmed and calibrated out without using a precision voltage source using a method similar to what the daily auto cal in the HP3458 does. It measures it's own internal 6+V reference. > >> > >> Every year or more, best to record the value of the nominal 6.2 volts, to measurement the ageing rate, by using a known external 10 volt standard. > >> To do that, an accurate external way is needed to find the exact ratio of the 6.2 volts to the external 10 volt standard. > >> This can be done by using a high resolution Linear DMM (aka HP3458A) OR a Kelvin-Varley divider (aka fluke 720 ) and null meter (aka Fluke 845). > >> Lets assume if one wants to do 0.1 PPM things, then they have the equipment and skill needed to measure a 10v to 6v ratio to 0.1 PPM. > >> After that ratio is known and recorded the units internal 10 volts can then be set to that same ratio without further need of the external 10V standard. > >> > >> The main advantage of all this extra trouble is low cost accuracy and stability. > >> There are no critical or expensive resistor needed or anything else critical, because all the TC and long term drift are canceled out whenever a manual trim is done using an external ratio device to check its own 10 V against it's own 6.2 volts. > >> > >> Nothing very special is needed to build this beside care and time. > >> To get the best performance, It does take a lot of Nut-time to match and select things. > >> Total parts cost with a well supplied analog junk box (or EBay) can be under $10. > >> > >> Any good 6V type reference device could be used. > >> I'm using a well aged and selected 1n825 because I have a lot of history with them and they work good without an oven. > >> I have not done any mailing test on these parts but they can be Zero TCed so that they do not change 0.1PPM over normal room temperature > >> and they show no change when powered down for short periods of time and tend to return to their original value when hit with a heat gun and cold stray, so it makes a good part to start experimenting with. > > -- > > John Devereux > > _______________________________________________ > 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.

Why is that ? I think it is correct. The opamp wants both its inputs at 6,2 V so he keeps his output around 5 V, so the output of the 7805 will be 10V. If the output drops the inverting input sees a lower voltage at the divider compared to the reference and the output raises until both are equal. The opamp is fed by the 10 V coming out the 7805. So far from its rail. I think it is a nice idea. But maybe I take a TL431 instead of the 7805 ( if i remember well that can deliver 100 mA and a better opamp like the OPA277. ) I think I gonna try one. If I'm not wrong Fluke does something like this in the 731A but I have the schematic not on hand right now. Fred PA4TIM Op 4 sep. 2011 om 18:28 heeft m k <m1k3k1@hotmail.com> het volgende geschreven: > > Hi, > The schematic has the opamp+ connected to the wrong point and there are some other details wrong. In another job there was an instrument that used an 1n829a run at about 7.5mA for a 10 V reference that used to hold to better than 1uV over lab temperatures. It was done using the Op amp to generate the 10V that directly drove the 7.5mA into the zener, select on test resistors. but for 0.1ppm you need to address opamp heating and unintended thermocouples everywhere. > >> From: john@devereux.me.uk >> To: volt-nuts@febo.com >> Date: Sun, 4 Sep 2011 15:56:00 +0100 >> Subject: Re: [volt-nuts] Sub PPM 10V transfer standard >> >> >> Hi Bill, Warren, >> >> Whats the matter with it? The schematic is clear enough to me, >> interesting and seems quite clever to this wannabe nut. >> >> If I understand correctly, the opamp adjusts the divided-down output via >> the regulator so that it matches the reference. Everything is powered >> from the 10V output, eliminating problems with variable input voltages >> and opamp PSRR. >> >> I would be interested to know: >> >> - Any suggestions for the wirewound resistors? Vishay bulk metal foil >> resistors are not cheap but say 5ppm wirewounds are not pennies >> either, I don't know how many you would need before a sufficient >> number could be matched. >> >> - I am unclear as to the precise function of the 160k and 40M >> resistors. I suppose one of them shifts the output, and the other >> compensates the zener so it still runs at the same minimum-tempco >> current? >> >> The zeners do look interesting as references, I admit I have been >> fixated on LTZ1000s. >> >> Farnell/Newark don't stock the 1N825 but they do stock the >> 1N825A. >> >> <http://uk.farnell.com/american-power-devices/1n825a/diode-zener-6-5v-0-25w-do35/dp/1651077?Ntt=1N825> >> >> According to its "datasheet" >> <http://www.farnell.com/datasheets/332215.pdf> there are also some >> similar looking parts that may have have improved tempco, for example >> 1N829A at 5ppm/K vs 1N825 at 20ppm/K. These are also available. There >> are even some 2ppm/K parts but these seem unobtainium. >> >> Are there any better datasheets? For example how does the tempco itself >> vary with temperature? The company does not seem to have a web presence. >> >> There >> >> Regards, >> >> John >> >> >> WB6BNQ <wb6bnq@cox.net> writes: >> >>> Warren, >>> >>> You need to do your schematics a little bit better. Besides, I think you attached the wrong one. There are a number of errors on the attached schematic. >>> >>> Bill....WB6BNQ >>> >>> WarrenS wrote: >>> >>>> Want a stable 10V ref? >>>> Best thing to do is get a fluke 732A/B and have it calibrated regularly for a few years. >>>> BUT >>>> For the Extreme (or cheap) Volt Nut, There is another choice. >>>> Attached is the circuit of a 0.1 PPM, 10 volt transfer standard. >>>> >>>> It has ONLY ONE critical part that limits its accuracy and stability. >>>> Everything else can be made to have so little effect as to be insignificant or else can be trimmed and calibrated out without using a precision voltage source using a method similar to what the daily auto cal in the HP3458 does. It measures it's own internal 6+V reference. >>>> >>>> Every year or more, best to record the value of the nominal 6.2 volts, to measurement the ageing rate, by using a known external 10 volt standard. >>>> To do that, an accurate external way is needed to find the exact ratio of the 6.2 volts to the external 10 volt standard. >>>> This can be done by using a high resolution Linear DMM (aka HP3458A) OR a Kelvin-Varley divider (aka fluke 720 ) and null meter (aka Fluke 845). >>>> Lets assume if one wants to do 0.1 PPM things, then they have the equipment and skill needed to measure a 10v to 6v ratio to 0.1 PPM. >>>> After that ratio is known and recorded the units internal 10 volts can then be set to that same ratio without further need of the external 10V standard. >>>> >>>> The main advantage of all this extra trouble is low cost accuracy and stability. >>>> There are no critical or expensive resistor needed or anything else critical, because all the TC and long term drift are canceled out whenever a manual trim is done using an external ratio device to check its own 10 V against it's own 6.2 volts. >>>> >>>> Nothing very special is needed to build this beside care and time. >>>> To get the best performance, It does take a lot of Nut-time to match and select things. >>>> Total parts cost with a well supplied analog junk box (or EBay) can be under $10. >>>> >>>> Any good 6V type reference device could be used. >>>> I'm using a well aged and selected 1n825 because I have a lot of history with them and they work good without an oven. >>>> I have not done any mailing test on these parts but they can be Zero TCed so that they do not change 0.1PPM over normal room temperature >>>> and they show no change when powered down for short periods of time and tend to return to their original value when hit with a heat gun and cold stray, so it makes a good part to start experimenting with. >> >> -- >> >> John Devereux >> >> _______________________________________________ >> 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.

In message <BLU160-W589E11CA52703B9DEA19E7911A0@phx.gbl>, m k writes: >But for 0.1ppm you need to address opamp heating and unintended >thermocouples everywhere. Actually, if you go that far anyway, adding a peltier and ntc to hold the temperature constant is not a big deal. Build the PCB (both Vref & Tctl) as surface-mount with no through-hole and attach it directly to the Peltier with a thermal pad The Peltier needs very little power if the temperature difference is on the order of a few degrees and the power dissipation in the controlled volume is low. When I ran the MAX6350 experiment, the 12V Peltier was running at slightly less than 1V most of the time. I would let the Tctl output an analog control voltage and put the Peltier power-driver outside the controlled volume. I would recommend preregulating the input power with a linear regulator (also outside the controlled volume) so that the power dissipation in the controlled volume is as constant as possible. -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 phk@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.

Hi, could you be more specific? m k <m1k3k1@hotmail.com> writes: > Hi, The schematic has the opamp+ connected to the wrong point and > there are some other details wrong. Do you mean the +input or the +supply? They both look OK to me but noone is saying what the problem is! > In another job there was an instrument that used an 1n829a run at > about 7.5mA for a 10 V reference that used to hold to better than 1uV > over lab temperatures. It was done using the Op amp to generate the > 10V that directly drove the 7.5mA into the zener, select on test > resistors. but for 0.1ppm you need to address opamp heating and > unintended thermocouples everywhere. This is essentially what he is doing as far as I can see. But the opamp heating is addressed by using the 78L05 as a "power driver", so the opamp generates no heat! The heat-generating part is taken off-chip (or off-board perhaps) so it does not affect the input offset voltage. I agree 0.1ppm seems optimistic, but then I haven't built it. In my admittedly limited experience thermocouples are not such a problem if you keep everything isothermal, i.e. avoid temperature gradients. Stating the obvious perhaps. Even if it were 1ppm instead 0.1ppm I think this is still an order of magnitude better than the homebrew ideas discussed so far (if achievable). John > >> From: john@devereux.me.uk >> To: volt-nuts@febo.com >> Date: Sun, 4 Sep 2011 15:56:00 +0100 >> Subject: Re: [volt-nuts] Sub PPM 10V transfer standard >> >> >> Hi Bill, Warren, >> >> Whats the matter with it? The schematic is clear enough to me, >> interesting and seems quite clever to this wannabe nut. >> >> If I understand correctly, the opamp adjusts the divided-down output via >> the regulator so that it matches the reference. Everything is powered >> from the 10V output, eliminating problems with variable input voltages >> and opamp PSRR. >> >> I would be interested to know: >> >> - Any suggestions for the wirewound resistors? Vishay bulk metal foil >> resistors are not cheap but say 5ppm wirewounds are not pennies >> either, I don't know how many you would need before a sufficient >> number could be matched. >> >> - I am unclear as to the precise function of the 160k and 40M >> resistors. I suppose one of them shifts the output, and the other >> compensates the zener so it still runs at the same minimum-tempco >> current? >> >> The zeners do look interesting as references, I admit I have been >> fixated on LTZ1000s. >> >> Farnell/Newark don't stock the 1N825 but they do stock the >> 1N825A. >> >> <http://uk.farnell.com/american-power-devices/1n825a/diode-zener-6-5v-0-25w-do35/dp/1651077?Ntt=1N825> >> >> According to its "datasheet" >> <http://www.farnell.com/datasheets/332215.pdf> there are also some >> similar looking parts that may have have improved tempco, for example >> 1N829A at 5ppm/K vs 1N825 at 20ppm/K. These are also available. There >> are even some 2ppm/K parts but these seem unobtainium. >> >> Are there any better datasheets? For example how does the tempco itself >> vary with temperature? The company does not seem to have a web presence. >> >> There >> >> Regards, >> >> John >> >> >> WB6BNQ <wb6bnq@cox.net> writes: >> >> > Warren, >> > >> > You need to do your schematics a little bit better. Besides, I >> > think you attached the wrong one. There are a number of errors on >> > the attached schematic. >> > >> > Bill....WB6BNQ >> > >> > WarrenS wrote: >> > >> >> Want a stable 10V ref? Best thing to do is get a fluke 732A/B and >> >> have it calibrated regularly for a few years. BUT For the Extreme >> >> (or cheap) Volt Nut, There is another choice. Attached is the >> >> circuit of a 0.1 PPM, 10 volt transfer standard. >> >> >> >> It has ONLY ONE critical part that limits its accuracy and >> >> stability. Everything else can be made to have so little effect >> >> as to be insignificant or else can be trimmed and calibrated out >> >> without using a precision voltage source using a method similar to >> >> what the daily auto cal in the HP3458 does. It measures it's own >> >> internal 6+V reference. >> >> >> >> Every year or more, best to record the value of the nominal 6.2 >> >> volts, to measurement the ageing rate, by using a known external >> >> 10 volt standard. To do that, an accurate external way is needed >> >> to find the exact ratio of the 6.2 volts to the external 10 volt >> >> standard. This can be done by using a high resolution Linear DMM >> >> (aka HP3458A) OR a Kelvin-Varley divider (aka fluke 720 ) and null >> >> meter (aka Fluke 845). Lets assume if one wants to do 0.1 PPM >> >> things, then they have the equipment and skill needed to measure a >> >> 10v to 6v ratio to 0.1 PPM. After that ratio is known and >> >> recorded the units internal 10 volts can then be set to that same >> >> ratio without further need of the external 10V standard. >> >> >> >> The main advantage of all this extra trouble is low cost accuracy >> >> and stability. There are no critical or expensive resistor needed >> >> or anything else critical, because all the TC and long term drift >> >> are canceled out whenever a manual trim is done using an external >> >> ratio device to check its own 10 V against it's own 6.2 volts. >> >> >> >> Nothing very special is needed to build this beside care and time. >> >> To get the best performance, It does take a lot of Nut-time to >> >> match and select things. Total parts cost with a well supplied >> >> analog junk box (or EBay) can be under $10. >> >> >> >> Any good 6V type reference device could be used. I'm using a well >> >> aged and selected 1n825 because I have a lot of history with them >> >> and they work good without an oven. I have not done any mailing >> >> test on these parts but they can be Zero TCed so that they do not >> >> change 0.1PPM over normal room temperature and they show no change >> >> when powered down for short periods of time and tend to return to >> >> their original value when hit with a heat gun and cold stray, so >> >> it makes a good part to start experimenting with. -- John Devereux

In message <871uvwp2xa.fsf@devereux.me.uk>, John Devereux writes: >Even if it were 1ppm instead 0.1ppm I think this is still an order of >magnitude better than the homebrew ideas discussed so far (if >achievable). The problem is, to be sure you need to build at least three :-) If somebody were to do a group-buy/kit of something like this, I would buy a handful... -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 phk@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.

"Poul-Henning Kamp" <phk@phk.freebsd.dk> writes: > In message <871uvwp2xa.fsf@devereux.me.uk>, John Devereux writes: > >>Even if it were 1ppm instead 0.1ppm I think this is still an order of >>magnitude better than the homebrew ideas discussed so far (if >>achievable). > > The problem is, to be sure you need to build at least three :-) > > If somebody were to do a group-buy/kit of something like this, > I would buy a handful... Well I've already ordered some zeners :) Need a lot more experience/tinkering though. But perhaps some "gold-plated" parts would do instead. I'm thinking something like Vishay VAR or Z201 series resistors instead of all the wirewounds, about 10 or 15 UKP each in singles, would need 3 or 4 I think. The OP07 is looking a bit long in the tooth, there are more modern parts or one could consider a chopper amp. But I do like the idea of seeing what can be done with low cost parts too. -- John Devereux

Hi, When looking closer at it the problem was the initial confusion caused by the guard rings as drawn. It was hard to be certain what should be connected where. When I drew it out for myself it all made sense. Keeping everything isothermal is always difficult. But one answer is always the medicinal parraffin. > From: john@devereux.me.uk > To: volt-nuts@febo.com > Date: Sun, 4 Sep 2011 19:39:45 +0100 > Subject: Re: [volt-nuts] Sub PPM 10V transfer standard > > > Hi, could you be more specific? > > m k <m1k3k1@hotmail.com> writes: > > > Hi, The schematic has the opamp+ connected to the wrong point and > > there are some other details wrong. > > Do you mean the +input or the +supply? They both look OK to me but noone > is saying what the problem is! > > > In another job there was an instrument that used an 1n829a run at > > about 7.5mA for a 10 V reference that used to hold to better than 1uV > > over lab temperatures. It was done using the Op amp to generate the > > 10V that directly drove the 7.5mA into the zener, select on test > > resistors. but for 0.1ppm you need to address opamp heating and > > unintended thermocouples everywhere. > > This is essentially what he is doing as far as I can see. But the opamp > heating is addressed by using the 78L05 as a "power driver", so the > opamp generates no heat! The heat-generating part is taken off-chip (or > off-board perhaps) so it does not affect the input offset voltage. > > I agree 0.1ppm seems optimistic, but then I haven't built it. In my > admittedly limited experience thermocouples are not such a problem if > you keep everything isothermal, i.e. avoid temperature > gradients. Stating the obvious perhaps. > > Even if it were 1ppm instead 0.1ppm I think this is still an order of > magnitude better than the homebrew ideas discussed so far (if > achievable). > > John > > > > >> From: john@devereux.me.uk > >> To: volt-nuts@febo.com > >> Date: Sun, 4 Sep 2011 15:56:00 +0100 > >> Subject: Re: [volt-nuts] Sub PPM 10V transfer standard > >> > >> > >> Hi Bill, Warren, > >> > >> Whats the matter with it? The schematic is clear enough to me, > >> interesting and seems quite clever to this wannabe nut. > >> > >> If I understand correctly, the opamp adjusts the divided-down output via > >> the regulator so that it matches the reference. Everything is powered > >> from the 10V output, eliminating problems with variable input voltages > >> and opamp PSRR. > >> > >> I would be interested to know: > >> > >> - Any suggestions for the wirewound resistors? Vishay bulk metal foil > >> resistors are not cheap but say 5ppm wirewounds are not pennies > >> either, I don't know how many you would need before a sufficient > >> number could be matched. > >> > >> - I am unclear as to the precise function of the 160k and 40M > >> resistors. I suppose one of them shifts the output, and the other > >> compensates the zener so it still runs at the same minimum-tempco > >> current? > >> > >> The zeners do look interesting as references, I admit I have been > >> fixated on LTZ1000s. > >> > >> Farnell/Newark don't stock the 1N825 but they do stock the > >> 1N825A. > >> > >> <http://uk.farnell.com/american-power-devices/1n825a/diode-zener-6-5v-0-25w-do35/dp/1651077?Ntt=1N825> > >> > >> According to its "datasheet" > >> <http://www.farnell.com/datasheets/332215.pdf> there are also some > >> similar looking parts that may have have improved tempco, for example > >> 1N829A at 5ppm/K vs 1N825 at 20ppm/K. These are also available. There > >> are even some 2ppm/K parts but these seem unobtainium. > >> > >> Are there any better datasheets? For example how does the tempco itself > >> vary with temperature? The company does not seem to have a web presence. > >> > >> There > >> > >> Regards, > >> > >> John > >> > >> > >> WB6BNQ <wb6bnq@cox.net> writes: > >> > >> > Warren, > >> > > >> > You need to do your schematics a little bit better. Besides, I > >> > think you attached the wrong one. There are a number of errors on > >> > the attached schematic. > >> > > >> > Bill....WB6BNQ > >> > > >> > WarrenS wrote: > >> > > >> >> Want a stable 10V ref? Best thing to do is get a fluke 732A/B and > >> >> have it calibrated regularly for a few years. BUT For the Extreme > >> >> (or cheap) Volt Nut, There is another choice. Attached is the > >> >> circuit of a 0.1 PPM, 10 volt transfer standard. > >> >> > >> >> It has ONLY ONE critical part that limits its accuracy and > >> >> stability. Everything else can be made to have so little effect > >> >> as to be insignificant or else can be trimmed and calibrated out > >> >> without using a precision voltage source using a method similar to > >> >> what the daily auto cal in the HP3458 does. It measures it's own > >> >> internal 6+V reference. > >> >> > >> >> Every year or more, best to record the value of the nominal 6.2 > >> >> volts, to measurement the ageing rate, by using a known external > >> >> 10 volt standard. To do that, an accurate external way is needed > >> >> to find the exact ratio of the 6.2 volts to the external 10 volt > >> >> standard. This can be done by using a high resolution Linear DMM > >> >> (aka HP3458A) OR a Kelvin-Varley divider (aka fluke 720 ) and null > >> >> meter (aka Fluke 845). Lets assume if one wants to do 0.1 PPM > >> >> things, then they have the equipment and skill needed to measure a > >> >> 10v to 6v ratio to 0.1 PPM. After that ratio is known and > >> >> recorded the units internal 10 volts can then be set to that same > >> >> ratio without further need of the external 10V standard. > >> >> > >> >> The main advantage of all this extra trouble is low cost accuracy > >> >> and stability. There are no critical or expensive resistor needed > >> >> or anything else critical, because all the TC and long term drift > >> >> are canceled out whenever a manual trim is done using an external > >> >> ratio device to check its own 10 V against it's own 6.2 volts. > >> >> > >> >> Nothing very special is needed to build this beside care and time. > >> >> To get the best performance, It does take a lot of Nut-time to > >> >> match and select things. Total parts cost with a well supplied > >> >> analog junk box (or EBay) can be under $10. > >> >> > >> >> Any good 6V type reference device could be used. I'm using a well > >> >> aged and selected 1n825 because I have a lot of history with them > >> >> and they work good without an oven. I have not done any mailing > >> >> test on these parts but they can be Zero TCed so that they do not > >> >> change 0.1PPM over normal room temperature and they show no change > >> >> when powered down for short periods of time and tend to return to > >> >> their original value when hit with a heat gun and cold stray, so > >> >> it makes a good part to start experimenting with. > > -- > > John Devereux > > _______________________________________________ > 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.