VoltCard design with LM399: does this look sane?

[technix]

This circuit is my design of a self-contained voltage reference card. It plugs into target circuit with a 7-pin connector. It takes a single 12V input rail and outputs a reference voltage. The output voltage, uncertainty, and a digital certificate of calibration is stored in the onboard EEPROM, which is living on a separate digital portion of the card. When this VoltCard is inserted into the target assembly the onboard microcontroller will read the data and control relays accordingly.

My board used OP07C which is all I have. Any pin compatible op amp can be used in this circuit so if you volt nuts want one of this VoltCard with enhanced precision I can get OP177 onto that footprint.

The two ten turn trim pots adjusts the constant current to the Zener, and the output voltage respectively. All other resistors in the analog section on this board are 0603 SMT resistors with 25ppm tempco (initial errors are trimmed out with the pots)

EDIT: Changed board design to a simpler LM317 constant current source.

The digital section in the bottom right corner is independent to the analog section.

[Andreas]

Hello,

just some annotations:

the trimmers (200ppm/K) will spoil your 1ppm/K TC of the reference.
(you could use a cheaper band gap reference for this).

The LM317 needs a minimum load current of 5-10mA so a 1mA current source is a illusion.

You should kelvin sense the output voltage of the LM399 otherwise the heater current will change output voltage.
(this will need a additional ground pin).

with best regards

Andreas

[technix]

Hello,

just some annotations:

the trimmers (200ppm/K) will spoil your 1ppm/K TC of the reference.
(you could use a cheaper band gap reference for this).

The LM317 needs a minimum load current of 5-10mA so a 1mA current source is a illusion.

You should kelvin sense the output voltage of the LM399 otherwise the heater current will change output voltage.
(this will need a additional ground pin).

with best regards

Andreas

If the trimmer will ruin my day, what is the correct way of doing this then? Tap the Vz off directly with the OP07C as a voltage follower, write the measured voltage into the EEPROM, and let the digital side of things to compensate?

If I bring the Zener current up to 15mA, will LM317 work?

[Vgkid]

The lm399 zener only needs an iZ of 1.0mA. I was contemplating using a 10v voltage to power it, with a stable 10k resistor in series to provide the 1ma of current.

[technix]

The lm399 zener only needs an iZ of 1.0mA. I was contemplating using a 10v voltage to power it, with a stable 10k resistor in series to provide the 1ma of current.

I want some sort of better constant current source so I used that LM317, and since LM399 can safely takes a 15mA current I am just pushing that current into it. The 12V was a choice of convenience as the final board will be powered using an ATX power supply where 12V is really a choice of convenience. This VoltCard is really a replaceable plug in board.

[kjs]

The lm399 zener only needs an iZ of 1.0mA. I was contemplating using a 10v voltage to power it, with a stable 10k resistor in series to provide the 1ma of current.

I want some sort of better constant current source so I used that LM317, and since LM399 can safely takes a 15mA current I am just pushing that current into it. The 12V was a choice of convenience as the final board will be powered using an ATX power supply where 12V is really a choice of convenience. This VoltCard is really a replaceable plug in board.

but the LM317 is a VERY noisy regulator and factors worse than the spec of the LM399. If you look at the LM399 datasheet the better solution is to use a real low noise and temperature stable voltage regulator and a low temp-co resistor.

[radioFlash]

I'm interested in this topic and found another thread on the matter of 10V LM399 references (see the link below). I'll be interested in what you end up with.

https://www.eevblog.com/forum/projects/lm399-based-10-v-reference/

[Vgkid]

This is the circuit that will be powering my gpsdo, instead of the lm317 it will use a lm338.
any critiques, realize I left out the transformer/rectifier/10000uF of filter caps.

[GK]

You can use a variation of this circuit from the LM399 datasheet to trim the reference voltage without being bothered by the temperature coefficient of the trimpot.

[Andreas]

I want some sort of better constant current source so I used that LM317, and since LM399 can safely takes a 15mA current I am just pushing that current into it. The 12V was a choice of convenience as the final board will be powered using an ATX power supply where 12V is really a choice of convenience. This VoltCard is really a replaceable plug in board.

Hello,

datasheet spec is 1mA +/- 0.1% for long term stability.
Higher currents will not be useful.

Best way to do this is a 3K resistor from 10V output to supply the zener.
(you might need a startup cirquit).
2nd best way is to use a 7K5 resistor to a stabilized 15V supply.

For trimming I recommend the scheme of AD587 datasheet.
(where the internal resistors of AD587 have a good T.C. tracking ).

I would not use a switchmode supply together with a precision device in the LM399 class.

With best regards

Andreas

[babysitter]

How is the digital side going to fix anything?
You throw a lot of transistors on the problem of specifying the voltage references behaviour - a EEPROM.
The digital side has something - relay-based KVM or switchable feedback resistor or silicon RDAC.
So, Voltcard voltage times 1,428571429 might give 10V - are you sure its not doing Voltcard * 1,428571428?
But aren't you just assuming that the digital controlled post-conditioning does what it is meant to do, or are you actually looking at the ratio of the reference to the output??

[cellularmitosis]

EDIT: Changed board design to a simpler LM317 constant current source.

Welcome to the path of becoming a volt nut!   

The last time I fooled around with the LM399 and read through the LM399 mega thread on this forum, if I recall correctly the stability of the LM399 voltage ended up being dominated by your ability to drive a stable current through it.  I can't remember the exact relationship off the top of my head, but if your current source drifts by X microamps, the 399 drifts by Y microvolts.  (I believe this means the LM317 is going to be the biggest source of error for your design).

The design presented in the LM399 mega thread ( https://www.eevblog.com/forum/projects/lm399-based-10-v-reference/ ) is a really interesting one, because it uses the LM399 itself to control how much current goes through it.  Its kind of hard to wrap your head around, so I drew up a simplified version of the schematic in a way which makes sense to me:



That circuit produces roughly 10 volts at the output, but you can simply change the two 3k resistors to whatever you want to get a different output voltage (e.g. 4k would give you 11 volts, 5k would give you 12 volts...).

The key thing to understand is that, as soon as there is any current at all going through the LM399, it is near 7V.  That causes the op amp to try to make its other input close to 7V, and the only way it can do that is to drive 1mA through the 7k resistor.  And once that happens, you are also driving (a very stable) 1mA through the LM399 as well.  Pretty neat!

Here's an idea I've been sitting on for a while (which is based on the above circuit) to create a mini transfer standard to calibrate the various ranges on my HP 3478A bench meter:

[Andreas]

EDIT: Changed board design to a simpler LM317 constant current source.

Welcome to the path of becoming a volt nut!   

I can't remember the exact relationship off the top of my head, but if your current source drifts by X microamps, the 399 drifts by Y microvolts.  (I believe this means the LM317 is going to be the biggest source of error for your design).

The design presented in the LM399 mega thread ( https://www.eevblog.com/forum/projects/lm399-based-10-v-reference/ ) is a really interesting one, because it uses the LM399 itself to control how much current goes through it.

Hello,

you bring it to the point. Thank you for the schematics.

The relationship is the about 0.5 to 1 Ohms dynamic resistance of the LM399.
So each uA will give around 1uV drift.

With best regards

Andreas

[GK]

EDIT: Changed board design to a simpler LM317 constant current source.

Welcome to the path of becoming a volt nut!   

I can't remember the exact relationship off the top of my head, but if your current source drifts by X microamps, the 399 drifts by Y microvolts.  (I believe this means the LM317 is going to be the biggest source of error for your design).

The design presented in the LM399 mega thread ( https://www.eevblog.com/forum/projects/lm399-based-10-v-reference/ ) is a really interesting one, because it uses the LM399 itself to control how much current goes through it.

Hello,

you bring it to the point. Thank you for the schematics.

The relationship is the about 0.5 to 1 Ohms dynamic resistance of the LM399.
So each uA will give around 1uV drift.

With best regards

Andreas

The temperature coefficient of the LM317 is 0.02% per deg_C, which is 200ppm. For a 1mA current source this works out to:

200*(1mA/1e6) = 200nA drift per deg_C.

With an effective DC resistance of 0.5 ohms, that will contribute an additional voltage drift of 100nV per deg_C.

The LM399's temperature coefficient is 0.5ppm, which works out to:

0.5*(6.95V/1e6) = 3.475uV drift per deg_C.

At almost 35 times less than the intrinsic instability of the LM399, the voltage drift term contributed by the LM317 current source is therefore relatively negligible.

I haven't done the sums but the noise contribution from the LM317 should be negligible as well.

[technix]

EDIT: Changed board design to a simpler LM317 constant current source.

Welcome to the path of becoming a volt nut!   

I can't remember the exact relationship off the top of my head, but if your current source drifts by X microamps, the 399 drifts by Y microvolts.  (I believe this means the LM317 is going to be the biggest source of error for your design).

The design presented in the LM399 mega thread ( https://www.eevblog.com/forum/projects/lm399-based-10-v-reference/ ) is a really interesting one, because it uses the LM399 itself to control how much current goes through it.

Hello,

you bring it to the point. Thank you for the schematics.

The relationship is the about 0.5 to 1 Ohms dynamic resistance of the LM399.
So each uA will give around 1uV drift.

With best regards

Andreas

The temperature coefficient of the LM317 is 0.02% per deg_C, which is 200ppm. For a 1mA current source this works out to:

200*(1mA/1e6) = 200nA drift per deg_C.

With an effective DC resistance of 0.5 ohms, that will contribute an additional voltage drift of 100nV per deg_C.

The LM399's temperature coefficient is 0.5ppm, which works out to:

0.5*(6.95V/1e6) = 3.475uV drift per deg_C.

At almost 35 times less than the intrinsic instability of the LM399, the voltage drift term contributed by the LM317 current source is therefore relatively negligible.

I haven't done the sums but the noise contribution from the LM317 should be negligible as well.

So I can roll with LM317 and normal 25ppm temp co 0603 SMT resistors and a Bourns ten turn pot tweaking the current without worrying about losing too many accuracy? That seem nice. And is it possible to match the resistors and LM317 to make their tempco cancel out?

[GK]

I haven't done the sums but the noise contribution from the LM317 should be negligible as well.

OK, I've just done some sums. TI's datasheet for the LM317 specifies noise in a 10Hz to 10kHz bandwidth to be 0.003% of Vout, or 30ppm. At Vout = 1.25V (the internal reference voltage) the total rms noise voltage in the 10Hz to 10kHz bandwidth is:

30*(1.25V/1e6) = 37.5uV.

To establish a 1mA constant current source with the LM317, a 1250 ohm resistor is required. The rms noise current will therefore be 37.5uV/1250 ohms = 30nA rms.

Form a graph in the LM399 datasheet, the reference diode has a typical dynamic resistance at 10Hz of 0.5 ohms, rising to 1 ohm at 1kHz and 3 ohms at 10kHz. I'll be really (overly) generous and say that the average dynamic resistance over the 10Hz-to-10kHz range is 2 ohms.

2 ohms resistance and 30nA of noise current gives a voltage noise contribution of 60nV rms in the specified 10Hz-10kHz bandwidth. The datasheet noise voltage spec for the LM399AH in a 10Hz-10kHz bandwidth is 7uV typical and 50uV max.

So, unless I've stuffed my basic sums up, the noise contribution of the LM317 current source is even less significant than its drift with temperature.

[GK]

So I can roll with LM317 and normal 25ppm temp co 0603 SMT resistors and a Bourns ten turn pot tweaking the current without worrying about losing too many accuracy? That seem nice. And is it possible to match the resistors and LM317 to make their tempco cancel out?

I wouldn't bother with a trimpot to tweak the current. The absolute value isn't important, just how much it drifts.

[technix]

EDIT: Changed board design to a simpler LM317 constant current source.

Welcome to the path of becoming a volt nut!   

The last time I fooled around with the LM399 and read through the LM399 mega thread on this forum, if I recall correctly the stability of the LM399 voltage ended up being dominated by your ability to drive a stable current through it.  I can't remember the exact relationship off the top of my head, but if your current source drifts by X microamps, the 399 drifts by Y microvolts.  (I believe this means the LM317 is going to be the biggest source of error for your design).

The design presented in the LM399 mega thread ( https://www.eevblog.com/forum/projects/lm399-based-10-v-reference/ ) is a really interesting one, because it uses the LM399 itself to control how much current goes through it.  Its kind of hard to wrap your head around, so I drew up a simplified version of the schematic in a way which makes sense to me:



That circuit produces roughly 10 volts at the output, but you can simply change the two 3k resistors to whatever you want to get a different output voltage (e.g. 4k would give you 11 volts, 5k would give you 12 volts...).

The key thing to understand is that, as soon as there is any current at all going through the LM399, it is near 7V.  That causes the op amp to try to make its other input close to 7V, and the only way it can do that is to drive 1mA through the 7k resistor.  And once that happens, you are also driving (a very stable) 1mA through the LM399 as well.  Pretty neat!

Here's an idea I've been sitting on for a while (which is based on the above circuit) to create a mini transfer standard to calibrate the various ranges on my HP 3478A bench meter:

I am investigating the LM317 constant current source issue and I remembered I had a few LP2985 regulators. Will those allow me to build a constant current source at 1mA? Or do I have to go down the path of Zener/current mirror?

[technix]

So I can roll with LM317 and normal 25ppm temp co 0603 SMT resistors and a Bourns ten turn pot tweaking the current without worrying about losing too many accuracy? That seem nice. And is it possible to match the resistors and LM317 to make their tempco cancel out?

I wouldn't bother with a trimpot to tweak the current. The absolute value isn't important, just how much it drifts.

So LM317 have trouble following up the 1mA current, how about this circuit, using TL431 and a current mirror?

[technix]

I haven't done the sums but the noise contribution from the LM317 should be negligible as well.

OK, I've just done some sums. TI's datasheet for the LM317 specifies noise in a 10Hz to 10kHz bandwidth to be 0.003% of Vout, or 30ppm. At Vout = 1.25V (the internal reference voltage) the total rms noise voltage in the 10Hz to 10kHz bandwidth is:

30*(1.25V/1e6) = 37.5uV.

To establish a 1mA constant current source with the LM317, a 1250 ohm resistor is required. The rms noise current will therefore be 37.5uV/1250 ohms = 30nA rms.

Form a graph in the LM399 datasheet, the reference diode has a typical dynamic resistance at 10Hz of 0.5 ohms, rising to 1 ohm at 1kHz and 3 ohms at 10kHz. I'll be really (overly) generous and say that the average dynamic resistance over the 10Hz-to-10kHz range is 2 ohms.

2 ohms resistance and 30nA of noise current gives a voltage noise contribution of 60nV rms in the specified 10Hz-10kHz bandwidth. The datasheet noise voltage spec for the LM399AH in a 10Hz-10kHz bandwidth is 7uV typical and 50uV max.

So, unless I've stuffed my basic sums up, the noise contribution of the LM317 current source is even less significant than its drift with temperature.

Also, don't bother too much on the sum now as the voltage follower op amp I used here is OP07C, not your fancy OP177. (Probably I will get a few of those later)

And about the "getting 10V out of this" situation if I am using the standard cell replacement circuit how do I calculate the resistors? And what kind of tempco I need to aim for? Will those 25ppm 0603 parts good enough?

[GK]

So LM317 have trouble following up the 1mA current, how about this circuit, using TL431 and a current mirror?

Dude, an LM317(L) will work fine here. No problem supplying 1mA. However if you want a better current source that's fine, just be aware of the very modest overall improvement in performance. However that said, I don't think that your TL431/PNP CC circuit shown there would work any better.

[technix]

So LM317 have trouble following up the 1mA current, how about this circuit, using TL431 and a current mirror?

Dude, an LM317(L) will work fine here. No problem supplying 1mA. However if you want a better current source that's fine, just be aware of the very modest overall improvement in performance. However that said, I don't think that your TL431/PNP CC circuit shown there would work any better.

LM317 seem have trouble reaching proper regulation at 1mA and that is why I am thinking about this TL431/PNP CC which will work with proper regulation at 1mA. Also this version consumes way less board space and I need to reclaim some board space to fit the trimable amplifier according to the LT app note (noticed the big blank area to the left of the op amp? That's where the precision amplification will happen)

[GK]

Oh, you're right, minimum load current is specified at 3.5mA typ, so that rules out an LM317, unless of course you just run a heavier quiescent current through the LM399 or shunt the excess with a parallel resistor.

[technix]

Oh, you're right, minimum load current is specified at 3.5mA typ, so that rules out an LM317, unless of course you just run a heavier quiescent current through the LM399 or shunt the excess with a parallel resistor.

Quiescent is 1mA or the noise can get out of control. So TL431/PNP CC it is. The space in the middle is going to be that amplifier but how should I choose the resistors?

[GK]

If Iq just has to be 1mA you can set the LM317 current to 5mA using the recommended 240 ohm resistor and put a 1k8 resistor (to shunt away 4mA) in parallel with the LM399.

As for choosing the resistors, etc, you'll just have to draw out your circuit and sit down with pen, paper and calculator and compute/determine the significance of the tolerances and drifts. For the most part it just boils down to ohms law and calculations of percentages along the lines of my prior posts. I have to call it quits for this evening.