How good are these parts if I build a voltage reference out of it?

[technix]

I am thinking about building a voltage reference, using TL431 as the voltage standard and OP07C as the precision amplifier. How good will that be?

And since I don't have those sexy decade switches, would using reed relays affect the precision? How to select the resistors and trim the voltage standard? Will the microcontroller controlling the relays affect the output, since the control to the reed relays obviously have to come from a string of 74HC595s on the microcontroller's SPI bus? Is the "dial in a voltage on a keypad and hit go" a good user experience?

One step further, since a microcontroller is called for already, does it make sense to make it computer controlled? Is there any existing protocol available?

[mzzj]

I am thinking about building a voltage reference, using TL431 as the voltage standard and OP07C as the precision amplifier. How good will that be?

And since I don't have those sexy decade switches, would using reed relays affect the precision? How to select the resistors and trim the voltage standard? Will the microcontroller controlling the relays affect the output, since the control to the reed relays obviously have to come from a string of 74HC595s on the microcontroller's SPI bus? Is the "dial in a voltage on a keypad and hit go" a good user experience?

One step further, since a microcontroller is called for already, does it make sense to make it computer controlled? Is there any existing protocol available?

Good or not depends what you need. TL431 would be about same level of accuracy as voltage standards 150 years ago 
There is probably about a metric ton of better voltage references than TL431.

[technix]

I am thinking about building a voltage reference, using TL431 as the voltage standard and OP07C as the precision amplifier. How good will that be?

And since I don't have those sexy decade switches, would using reed relays affect the precision? How to select the resistors and trim the voltage standard? Will the microcontroller controlling the relays affect the output, since the control to the reed relays obviously have to come from a string of 74HC595s on the microcontroller's SPI bus? Is the "dial in a voltage on a keypad and hit go" a good user experience?

One step further, since a microcontroller is called for already, does it make sense to make it computer controlled? Is there any existing protocol available?

Good or not depends what you need. TL431 would be about same level of accuracy as voltage standards 150 years ago 
There is probably about a metric ton of better voltage references than TL431.

Please suggest a part. I have checked recently that even from Shenzhen suppliers selling parts salvaged from other boards, LM399 and LTZ1000 are expensive like hell, costing US$10 and US$50 each.

[DanielS]

Please suggest a part. I have checked recently that even from Shenzhen suppliers selling parts salvaged from other boards, LM399 and LTZ1000 are expensive like hell, costing US$10 and US$50 each.

The whole point of having a voltage reference to use as a standard is stability. You are not going to get anything standard-worthy out of a 431. Unfortunately, better things usually come with a heftier price tag. There are reasons why the 399 and LTZ are such well-established go-to reference parts.

[mzzj]

Please suggest a part. I have checked recently that even from Shenzhen suppliers selling parts salvaged from other boards, LM399 and LTZ1000 are expensive like hell, costing US$10 and US$50 each.

REF02 is pretty common and you should be able to find some for 1-2usd/piece or less. Even brand new should be like 3-5 usd.
It's nowhere as good as LM399 or LTZ but much much better than TL431

Some others that you might find: ADR4550, REF192, MAX6050

[technix]

Please suggest a part. I have checked recently that even from Shenzhen suppliers selling parts salvaged from other boards, LM399 and LTZ1000 are expensive like hell, costing US$10 and US$50 each.

REF02 is pretty common and you should be able to find some for 1-2usd/piece or less. Even brand new should be like 3-5 usd.
It's nowhere as good as LM399 or LTZ but much much better than TL431

Some others that you might find: ADR4550, REF192, MAX6050

How good is MC1403? The datasheet quotes a 10ppm tempco.

[c4757p]

If you can't spend $10 on an LM399, you probably don't need a voltage reference at all.

[mzzj]

Please suggest a part. I have checked recently that even from Shenzhen suppliers selling parts salvaged from other boards, LM399 and LTZ1000 are expensive like hell, costing US$10 and US$50 each.

REF02 is pretty common and you should be able to find some for 1-2usd/piece or less. Even brand new should be like 3-5 usd.
It's nowhere as good as LM399 or LTZ but much much better than TL431

Some others that you might find: ADR4550, REF192, MAX6050

How good is MC1403? The datasheet quotes a 10ppm tempco.

Well its something like 10 to 100 times better than TL431 so a definite improvement.
Probably good enough to start playing around. If I remember correctly your workbench didn't include  fancy 8,5 digit meters or anything like that. 

[technix]

If you can't spend $10 on an LM399, you probably don't need a voltage reference at all.

I don't have a 8.5 digit DMM to justify that - I just need something that can be used to calibrate my 3.5 digit meters, sort of a crude version of a transfer standard that I will send to a calibration lab.

[technix]

Please suggest a part. I have checked recently that even from Shenzhen suppliers selling parts salvaged from other boards, LM399 and LTZ1000 are expensive like hell, costing US$10 and US$50 each.

REF02 is pretty common and you should be able to find some for 1-2usd/piece or less. Even brand new should be like 3-5 usd.
It's nowhere as good as LM399 or LTZ but much much better than TL431

Some others that you might find: ADR4550, REF192, MAX6050

How good is MC1403? The datasheet quotes a 10ppm tempco.

Well its something like 10 to 100 times better than TL431 so a definite improvement.
Probably good enough to start playing around. If I remember correctly your workbench didn't include  fancy 8,5 digit meters or anything like that. 

No all I have is 3.5 digit old ones. If I am to send one down to a calibration lab, should I finalize the design and send down that digitally controlled voltage standard assembly, or just the reference chip itself? It is just a working standard I will be used to calibrate my gear against.

[mzzj]

Please suggest a part. I have checked recently that even from Shenzhen suppliers selling parts salvaged from other boards, LM399 and LTZ1000 are expensive like hell, costing US$10 and US$50 each.

REF02 is pretty common and you should be able to find some for 1-2usd/piece or less. Even brand new should be like 3-5 usd.
It's nowhere as good as LM399 or LTZ but much much better than TL431

Some others that you might find: ADR4550, REF192, MAX6050

How good is MC1403? The datasheet quotes a 10ppm tempco.

Well its something like 10 to 100 times better than TL431 so a definite improvement.
Probably good enough to start playing around. If I remember correctly your workbench didn't include  fancy 8,5 digit meters or anything like that. 

No all I have is 3.5 digit old ones. If I am to send one down to a calibration lab, should I finalize the design and send down that digitally controlled voltage standard assembly, or just the reference chip itself? It is just a working standard I will be used to calibrate my gear against.

Many of the listed references are available with 0,02% accuracy. That should be more than enough to check your meters even without sending to calibration lab.

[technix]

Please suggest a part. I have checked recently that even from Shenzhen suppliers selling parts salvaged from other boards, LM399 and LTZ1000 are expensive like hell, costing US$10 and US$50 each.

REF02 is pretty common and you should be able to find some for 1-2usd/piece or less. Even brand new should be like 3-5 usd.
It's nowhere as good as LM399 or LTZ but much much better than TL431

Some others that you might find: ADR4550, REF192, MAX6050

How good is MC1403? The datasheet quotes a 10ppm tempco.

Well its something like 10 to 100 times better than TL431 so a definite improvement.
Probably good enough to start playing around. If I remember correctly your workbench didn't include  fancy 8,5 digit meters or anything like that. 

No all I have is 3.5 digit old ones. If I am to send one down to a calibration lab, should I finalize the design and send down that digitally controlled voltage standard assembly, or just the reference chip itself? It is just a working standard I will be used to calibrate my gear against.

Many of the listed references are available with 0,02% accuracy. That should be more than enough to check your meters even without sending to calibration lab.

I want to test the low voltage ranges too so I will need that accuracy and I probably will have to build an adjustable supply out of a reference chip, and that (as well as my old meters) will have to take a trip to the cal lab.

[ebastler]

I want to test the low voltage ranges too so I will need that accuracy and I probably will have to build an adjustable supply out of a reference chip, and that (as well as my old meters) will have to take a trip to the cal lab.

If you feel that you need calibration (repeatedly, if you are really serious about it?), the few extra $ for a better part should be negligible compared to the overall cost of the module and calibration service, shouldn't they?

[technix]

I want to test the low voltage ranges too so I will need that accuracy and I probably will have to build an adjustable supply out of a reference chip, and that (as well as my old meters) will have to take a trip to the cal lab.

If you feel that you need calibration (repeatedly, if you are really serious about it?), the few extra $ for a better part should be negligible compared to the overall cost of the module and calibration service, shouldn't they?

Calibrating a piece of gear with LTZ1000 at its heart here in Shanghai at a government-operated cal lab (cheapest option actually) cost the same as the LTZ1000 itself. If I used MC1403 probably it won't worth a trip down a cal lab, but I don't have equipment that justifies a purchase of LM399 or LTZ1000.

Maybe when I decided to get a 8.5-digit DMM I will need that. However before that, are you suggesting that the accuracy of MC1403 as posted in the datasheet is more than good enough for me to calibrate my gear against?

[ebastler]

Maybe when I decided to get a 8.5-digit DMM I will need that. However before that, are you suggesting that the accuracy of MC1403 as posted in the datasheet is more than good enough for me to calibrate my gear against?

The MC1403 is only specified at +- 1% absolute accuracy, it seems. So if you want an "absolute" standard, rather than just checking your DMMs for drift, that may not be enough for you. Some of the other references mentioned in the thread fare a factor of 10 better in this respect, e.g. the REF192 or the MAX6050 at 0.1% and 0.2%, respectively. At first glance, I have not seen one with the 0.02% accuracy mentioned by mzzj, however.

[macboy]

Maybe when I decided to get a 8.5-digit DMM I will need that. However before that, are you suggesting that the accuracy of MC1403 as posted in the datasheet is more than good enough for me to calibrate my gear against?

The MC1403 is only specified at +- 1% absolute accuracy, it seems. So if you want an "absolute" standard, rather than just checking your DMMs for drift, that may not be enough for you. Some of the other references mentioned in the thread fare a factor of 10 better in this respect, e.g. the REF192 or the MAX6050 at 0.1% and 0.2%, respectively. At first glance, I have not seen one with the 0.02% accuracy mentioned by mzzj, however.

REF102 is 0.025% and has very good stability over temperature and time. It is a good candidate to build a 3.5 digit standard that doesn't need any initial calibration, just build and use. If you add trimming capability and calibrate it, you can get much better accuracy that should hold true for a long time.

[edavid]

How about an AD588AQ?  Judging from eBay, there are plenty of them in China.

[technix]

Maybe when I decided to get a 8.5-digit DMM I will need that. However before that, are you suggesting that the accuracy of MC1403 as posted in the datasheet is more than good enough for me to calibrate my gear against?

The MC1403 is only specified at +- 1% absolute accuracy, it seems. So if you want an "absolute" standard, rather than just checking your DMMs for drift, that may not be enough for you. Some of the other references mentioned in the thread fare a factor of 10 better in this respect, e.g. the REF192 or the MAX6050 at 0.1% and 0.2%, respectively. At first glance, I have not seen one with the 0.02% accuracy mentioned by mzzj, however.

That is why I said I was thinking about sending a MC1403 down to a cal lab - to get an accurate initial reading. Then the MC1403 will have a good enough drift and tempco to allow me continuously calibrate my gear against.

[technix]

Maybe when I decided to get a 8.5-digit DMM I will need that. However before that, are you suggesting that the accuracy of MC1403 as posted in the datasheet is more than good enough for me to calibrate my gear against?

The MC1403 is only specified at +- 1% absolute accuracy, it seems. So if you want an "absolute" standard, rather than just checking your DMMs for drift, that may not be enough for you. Some of the other references mentioned in the thread fare a factor of 10 better in this respect, e.g. the REF192 or the MAX6050 at 0.1% and 0.2%, respectively. At first glance, I have not seen one with the 0.02% accuracy mentioned by mzzj, however.

REF102 is 0.025% and has very good stability over temperature and time. It is a good candidate to build a 3.5 digit standard that doesn't need any initial calibration, just build and use. If you add trimming capability and calibrate it, you can get much better accuracy that should hold true for a long time.

If I want more accuracy how many digits can I squeeze out of a REF102 with calibration capabilities? And how to design calibratable relay operated decade switches?

[icon]

Have a look at this chap's channel on Youtube - might be useful.

https://www.youtube.com/channel/UCDqryeq1kMDSEQwltWqASrA

John

[Andreas]

isl21009bfb850 : 0.01%  @ 5V
max6350: 0.02% @ 5V
ref102c: 0.025% @ 10V

With best regards

Andreas

[technix]

isl21009bfb850 : 0.01%  @ 5V
max6350: 0.02% @ 5V
ref102c: 0.025% @ 10V

With best regards

Andreas

Part #	Nominal voltage	Initial accuracy	Tempco	Drift	Price (from Shenzhen, unit: CNY)
ISL21009BFB825	2.5V	+/- 0.5mV (+/- 0.02%)	3ppm	50ppm/1000hrs	15/each
REF02AU	5V	+/- 0.015V (+/- 0.3%)	4ppm	100ppm/1000hrs	5/each
MC1403	2.5V	+/- 25mV (+/- 1%)	10ppm	No data	4.25/each (8.5/pack of 2)
MAX6350CPA	5V	+/- 1mV (+/- 0.02%)	1ppm	30ppm/1000hr	30/each

Maybe MC1403 is a bad buy...

I think I will get a few REF02 and start building from that. Hope OP07C is a good enough op amp for instrumentation purposes and I am still worried about the reed relays and a long string of 74HC595 used as decade switch replacement. What kind of power supply should I use? Is a maximum rating of 20V enough? Is punching in a voltage on a keypad and hit go a good user interface?

[technix]

Maybe when I decided to get a 8.5-digit DMM I will need that. However before that, are you suggesting that the accuracy of MC1403 as posted in the datasheet is more than good enough for me to calibrate my gear against?

The MC1403 is only specified at +- 1% absolute accuracy, it seems. So if you want an "absolute" standard, rather than just checking your DMMs for drift, that may not be enough for you. Some of the other references mentioned in the thread fare a factor of 10 better in this respect, e.g. the REF192 or the MAX6050 at 0.1% and 0.2%, respectively. At first glance, I have not seen one with the 0.02% accuracy mentioned by mzzj, however.

REF102 is 0.025% and has very good stability over temperature and time. It is a good candidate to build a 3.5 digit standard that doesn't need any initial calibration, just build and use. If you add trimming capability and calibrate it, you can get much better accuracy that should hold true for a long time.

Scored a few used LM399 at US$2.5/each, end of discussion. I am going to design my standard project around these.

[ebastler]

Scored a few used LM399 at US$2.5/each, end of discussion. I am going to design my standard project around these.

Congratulations -- good to see that your project is making progress. With that choice it will make sense to have your voltage reference calibrated once it's built, since the LM399 has great stability, but a rather large 2% initial tolerance.

[technix]

Scored a few used LM399 at US$2.5/each, end of discussion. I am going to design my standard project around these.

Congratulations -- good to see that your project is making progress. With that choice it will make sense to have your voltage reference calibrated once it's built, since the LM399 has great stability, but a rather large 2% initial tolerance.

So your suggestion here is go ahead and build the voltage reference out of LM399, and send it on a trip to the cal lab? Will its status being a used part affect its accuracy?