LM399 based 10 V reference

[Andreas]

Do i need a separate reference output for the DAC (REF which is a bufferd output of the LM3999) or could i use the 10V output. I thought i build a Voltage divider(switched Capacitor with an LTC1043) to feed the DAC.

The 9k and the 20K resistors need to be low TC ones. The 100 ohms and the 1k Resistor aren't so critical is this assumption right?

Is it better to use one 14 or 16 bit DAC or two (cheaper) 8 bit DACs.

Of course you can use the 10V output. But what happens to your DAC if anyone tries to feed back some voltage on the output? Or what happens when RF-noise is demodulated at the protection diodes of the DAC input? So it depends on your environment.

If a resistor contributes only 1% to the output voltage the TC can be a factor 100 larger than the one of the output voltage determining resistors. All in all you will want to maintain the (around room temperature usually better as) 1ppm/K spec of the LM399.

Usually the long term stability and the TC are not specified for 8 Bit DACs. If you are only trimming a few ppm´s with the DACs and do the main trimming by other resistors in parallel to the 9K + 20K  the 8 Bit devices may be sufficient.

With best regards

Andreas

[SimonSatCom]

Hi

Here are some pictures of my LM299 reference. I had about 50 of these LM299 in my junk box and figure with not do something with them.

http://www.simonthenerd.com/10Vref.htm

If I did this project today I would have changed some of the design. I would not have used an LT1013 to amplify the reference voltage to 10V. A chopper op-amp would have been better. I would probably have build the whole thing into an oven. But all these considerations have let me to do a LTZ1000A reference instead.

[branadic]

I had about 50 of these LM299 in my junk box and figure with not do something with them.

I wish I'd had such a junk box too

About the oven, there are second hand but cheap double oven stabilized OCXOs such as MV89A out there.

Pictures of the inside: http://www.rbarrios.com/projects/MV89A/

One could disassemble the XO circuit and assemble a reference circuit inside. This way the reference is shielded too and if the whole thing is soldered under inert gas condition humidity is also no deal anymore. Could be worth a try.
On the other hand it would be sad about the OCXO itself. You can't have everything

[branadic]

All the discussion, also the current one in the LTZ thread, leads me to the point:

Use a LMx99 with buffered output, feed the output into a LTC1043 based divide by 1.5V circuit to supply a pwm-dac with 30bit as shown by Andreas (based on the EDN publication) and feedback the output of the dac to the buffer of the reference buffer to realize a trimmable 10V reference output voltage.
This way the gain setting resistors can be fixed values and the complete arrangement is somewhat cheap to realize.

[Andreas]

Hello branadic,

thats why I´m focused on the LM399 at the moment.
I am ageing 16 LM399s to select the best one.
About half of them is already below 1ppm/khr.

First step then is a ADC with LM399 reference.

With best regards

Andreas

[branadic]

This is interesting to know, maybe you want to tell more about what you are planing to do (circuit desciption)?

Why LM399 and not LM199AH/883 with higher temperature specs?

http://www.ebay.de/itm/NS-ST-LM199AH-883-CAN4-Voltage-Reference-/221365837650?pt=Bauteile&hash=item338a6e9f52

[Andreas]

This is interesting to know, maybe you want to tell more about what you are planing to do (circuit desciption)?

Why LM399 and not LM199AH/883 with higher temperature specs?

Up to now:
- LM399 as reference (I simply have them already)
- LTC1043 as divider (*3/4 = 5.2 V reference)
- LTC2400 as ADC input range 10V with 2nd LTC1043 in 2:1 configuration
- processor PIC24FV32KA302 (but im running out of pins so this is subject to change)
- 2*16 digit LC-display    6,5 digits resolution (7,5 digits in slow mode with filtering)
- isolated RS232 interface
- rotary encoder with switch for adjustments
- battery powered up to 48 hrs with 12 NiMH cells
- plastic case with clear lid for display
- optional inner shielding

With best regards

Andreas

[Mickle T.]

Now I'm working on the similar project:
- 3x LM399H as reference. I have a 10-th of them (well aged). 3 have a long-term drift less than 1 ppm/year (~0.3 ppm/sqrt(kHr) )
- ADS1282 in oven
- Input amplifier (like in Sol. 7081) with bootstrap power source
- Compensated metal-foil divider (1:5)
- Plastic case with atmega, LCD e.t.c.

[Andreas]

Ok,

this will give the high end variant of my design 

Is that a metal plate on the right side within the housing for shielding?

With best regards

Andreas

[Mickle T.]

Yes, the metal plate on the right side is guard screen.

[SimonSatCom]

Looks like a nice project Mickle T.

- Simon

[Galaxyrise]

- LTC1043 as divider (*3/4 = 5.2 V reference)

Is that one or two 1043s? If you got 3/4 with just one IC, I have a puzzle to solve   If two, how did you opt to join the *3 and /4 stages? 

I tried to get 10/7 working in LTSpice, but I couldn't get the sim to be stable. I needed 5 ICs and some of them had to switch at the same time; it seemed like a mess to implement in reality.  It had the nice side effect of providing 1, 2, 5, 7, and 10V references, though.

[branadic]

Is that one or two 1043s? If you got 3/4 with just one IC, I have a puzzle to solve   If two, how did you opt to join the *3 and /4 stages? 

One 1043, if you stuck have a look on the solution here:

https://www.eevblog.com/forum/projects/building-a-7-decade-voltage-calibrator/msg304819/#msg304819

[Andreas]

Is that one or two 1043s? If you got 3/4 with just one IC, I have a puzzle to solve   If two, how did you opt to join the *3 and /4 stages? 

I tried to get 10/7 working in LTSpice, but I couldn't get the sim to be stable.

Damned: I should have patended my single LTC1043 *3/4 solution.

With best regards

Andreas

[branadic]

Damned: I should have patended my single LTC1043 *3/4 solution

Maybe you should have. 

I'm nearly finished with the schematics for the adjustable LMx99 reference. You could also say it's some kind of calibrator or transfer standard.

- LMx99 (6.75V ... 6.95V ... 7.3V) is tapped and divided by 3/4 (5.0625V ... 5.2125 ... 5.475V) using LTC1043
- the second tap goes into a LTC2057 booster with gain 1.25 (Z201 5k, 20k)
- the voltage of the divider is feed into the 30bit pwm-dac (modifications given by Andreas)
- the output of the pwm-dac is feed through the 20k resistor into the LTC2057 booster

This gives a digital adjustable reference using standard of the shelf Z201 resistor values.
You could use a DAC such as AD5791 instead, but this is much to expensive.

[Andreas]

Hello branadic,

Have I understood it right?

you have created a cirquit which is adjustable from around 8.2 to (8.6 + 5.2) V

I thought myself of a similar cirquit.
But with a single resistor divider.
The MAX4053A can be used up to a supply of +/- 8V which is sufficient for a 7V reference. (the LTC1152 with >= +/- 10V)
The output from 0..7V could be fed into a *1.5 amplifier (I thought of 1 or 2 DSMZ resistor dividers) giving 0 .. 10.5V

With best regards

Andreas

[branadic]

Hi Andreas,

I have drawn a circuit with LMx99 output voltage multiplied by 1.25 and the possibility to adjust that voltage with the pwm-dac to 10.xxxxxxV. Up to know it is just my documentation of an idea and not yet perfect.

branadic

[SimonSatCom]

Hi

In the datasheet for the LM399 (LM199, LM299) there is an application called "Portable Calibrator" (Figure 22 in the datasheet from TI).

http://www.ti.com/lit/ds/symlink/lm399.pdf

The circuit uses a 200k resistor in series with the reference zener diode. Does anyone know why they used 200k which would set the zener current between 25 uA and 55 uA. In the datasheet they say that the operating current is between 500 uA to 10 mA. Why do they then set the current in their application for a much lower value?

- Simon

[SimonSatCom]

Ah...well it may be the combined current for the reference and the heater...probably.

- Simon

[Galaxyrise]

The circuit uses a 200k resistor in series with the reference zener diode. Does anyone know why they used 200k which would set the zener current between 25 uA and 55 uA.

That 200k isn't setting the zener current; the opamp output is driving the zener through a 5k resistor.

[SimonSatCom]

Ahh, thanks for the input.

[Andreas]

Hello,

the 200K is only for start up in case the op-amp has a negative input offset voltage.
I also use a similar resistor in my LTZ1000A circuit.

With best regards

Andreas

[SimonSatCom]

Hi

Okay...so you did not use the "standard" application note schematic in the LT1000(A) datasheet. That sounds interesting since most people stick to that circuit. Cool.

[Andreas]

Hello,

you will find this pull-up resistor in the HP3458A reference board circuit.
Also the Datron reference circuit uses a pull up.
The LM399 too.
So probably in the datasheet of the LTZ they simply forgot to add one.

With best regards

Andreas

[SeanB]

Or in the LT test unit the opamp had a positive offset and was thus self starting. Get one that is not so and you will have it sitting stably at zero a lot of the time until noise turns it on.