LM399 based 10 V reference

[miro123]

The 100nF ser 10ohm - EricM indicated the new ADR1399 is more sensitive to capacitive loads than the LM399..

Where can I find the ADR1399 datasheet?

[Kleinstein]

AFAIK there is not yet a datasheet for the ADR1399. There was a more or less unoffcial anouncment that there will be one coming out late this year. Suggestest changes to the LM399 were:
4 x the zener current  (so likely more like 2 mA to run instead of some 1 mA as the usual LM399 current)
1/2 the noise
lower differential output impedance  (so less critical resistor for the current, is already not that critical)
more picky about capacitance at the ouput, but no details given.
Likely similar case, so it may work as a drop in with minimal changes (more current and maybe ESR for the cap if present).

A voltage regualtor is always a bit tricky with a low ESR cap at the output. A voltage regulating loop tends to have an ouput impedance that goes up with frequency and thus means an inductive looking impedance. Capacitance with ESR is usually easier to regulate. We don't know how much ESR is needed, but for the PCB layout it is definitely a good idea to have a resistor planed. One can still change to 10 Ohms to 1 Ohms or 100 Ohms if needed. I doublt they would need a very low ESR - this usually needs extra effort to get stable.

[miro123]

4 x the zener current  (so likely more like 2 mA to run instead of some 1 mA as the usual LM399 current)
1/2 the noise
lower differential output impedance  (so less critical resistor for the current, is already not that critical)

Thanks  for sharing.  I have another question
Twice the current means 50% reduction of bootstrap resistor. e.g. from 3k -> 1,5k
According to me - twice the current and half the Zenner dynamic resistance requires the same resistor tolerance. Am I right?
E.g LM399 has 1 Ohm dyn resistance and Rb=3k . Hypothetic ADR1399 has 0,5 Ohm & Rb=1,5k. Both circuit will require the same resistor ppms.

[iMo]

There is none free zener in the 399. The zener is included in a regulator, consisting of several transistors, resistors, caps etc.
Afaik the zener current in LM399 is set to 250uA.

[Cerebus]

Afaik the zener current in LM399 is set to 250uA.

Yup.

Current through D3 is held constant at 250 μA by a 2k resistor across the emitter base of Q13 while the emitter-base voltage of Q13 nominally temperature compensates the reference voltage.

[iMo]

I collected several MAC199 (LM399 clone in TO18 case) with broken heater. I've tried to utilize the parasitic diode for chip temperature measurement.

Basically it works, with say 80uA fw current the voltage at the heater's ground changes like (aprox)

31C (fingers)  620mV
26C (amb)     637mV
IPA large drop  677mV

There is an issue with it, however - while the 80uA current is injected into the heater's gnd (parasitic diode anode) the Vref (standard 10V Vref like the above) increases by +3000ppm. That is way more than a drop on the bond wire - thus it messes up something on the chip, it seems..
PS: or chip temperature increase?? That is 50uW loss at the diode..

The "cold zener" itself shows something like 30-40ppm/K.

It seems it could be useful for measuring the chip temperature before the Vref measurement, then the diode current has to be switched off (a mechanical switch for example) and you may measure the Vref.
The loss at the diode during the measurement is aprox 13uW with 20uA.

[Kleinstein]

AFAIK the neg. side of the heat part is the substrate and should be the most negative voltage in the whole system. A current troub a parasitic diode can injsct substarte current and this can upset the rest of the circuit electrical  (current from the substrate to parts of the circuit).

The MAX199 with a broken heater is essentially like an LM129, so a metal case but no heater. If the TC is reasonably low, one may get away with an external oven. There is no absolute need for a sensor directly at the chip. With a constant power at the reference it is OK to measure outside.  An external oven is slower to stabilize, but not that bad.

[iMo]

Sure, that oven is the standard approach..
Back to the roots - what would be the max resistor value in the RC filter (from the 399 zener to the opmap's input) for OPxxx, or, LTC2057?
The Johnson noise is 0.12uV rms with 80k at 33C, dF 10Hz..
A single 399 has something like 1-2uV rms.
We should add opamp's current and voltage noise, right?

PS: below LTC2057 (89nV rms) and OP177 (387nV rms) noise in 0.1 to 10Hz, 10k resistor. Not sure it shows all contributions..

[Kleinstein]

With a RC filter you have the OPs input current (bias and curent noise) and the Johnson noise of the resistor.
With a reasonable size capacitor (e.g. 10 µF) the upper frequency limit would be well below 10 Hz. So the relevant noise is not 0.1 to 10 Hz, more like 0.01 to 1 Hz or even lower. The main parts there are the 1/f noise of the LM399 (or similar) and the 1/f noise of the OP. Especially the current noise may have quite some 1/f part.
So the OP07 would not be a good choice for high resistance (e.g. > 30 K). More like OPA202 as a BJT based one with low current noise, or an AZ OP like LTC2057, AD8628 or maybe max4238. The specs for the current noise of AZ OPs seem to be not very relable - 1 would not count on all the numbers there, especially when very low.  I would consider some 100 K a reasonable upper limit as than one starts to have to worry about leakage curents in the pA range that may change. The main noise source would still be the LM399, even if using 4 or so.  Still filtering the reference is limited. The filter will also need quite some settling. It is not only the RC time constant, but there are also slow DA effects that need to settle when it comes to settling to something like 1 ppm of the final value. This part starts smaller but can be 100 or 1000 times slower.

The capacitor may also need a reasonable stable temperature (e.g. thermal shield): I can see a small thermal effect even with only some 5 K and 4.7 µF (MKS with 3 x muliplication), it would get more with higher resistance.

If it is about a more pulsed operation (e.g. 99 ms zener and 1 ms temperature measurement) the more suitable way would be a switch to turn off the connection when in temperature mode.

[dietert1]

Those considerations are subject to experimental confirmation. Caps to be used for filtering references need to be qualified to the same level of 1 ppm or below. Don't assume anything.

I made a PWM 10V stage for a 5x LM399 array (schematic above). In my case the filter is the PWM filter and it is second order with a bandwidth less than 1 Hz. Resistors used are 22K +47K. From the very start i implemented a bootstrap scheme with electrolytic caps to keep the voltage over the filter caps near zero. But the electrolytic caps are very temperature sensitive. When i ran the unit in an incubator, noise due to forced air flow went up to 2.5 ppm p2p, while average stability was near perfect. When wrapped in tissue, noise is down to about 0.3 ppm p2p, but the unit exhibits notable TC, as it partially decouples from incubator temperature. The next revision will be without bootstrap.

Also i remember a leakage test with a 650 uF foil capacitor (AVX FFLI6L0657K). I could see temperature variations as voltage changes - over several hours. I think when capacitance changes with temperature, the constant charge in the cap will result in changing voltages. It was difficult to model as the capacitor has temperature gradients inside.

Regards, Dieter

[iMo]

..
If it is about a more pulsed operation (e.g. 99 ms zener and 1 ms temperature measurement) the more suitable way would be a switch to turn off the connection when in temperature mode.

You can even use the substrate diode as a "heater", imho. Say 20mA*0.65V=130mW. TO18 thermal resistance is something like 300C/Watt, 0.13*300 = 39C, with Tamb=25 you'll get Tpackage=25+39=64C.
With 50mA diode current you may anneal all defects off the silicon 

[Kleinstein]

Current through the substrate diode can interact with the rest of the circuit. Usually this is not in a good way. So once there is substarte current all bets are off.



For filtering I would not consider electrolytic caps practical, because of the temperature effect and leakage that can change with time and temperature. The DA will also requite extremely long settling - maybe not as long as the LM399, but it would start all over after power off.
The question is more if polyester caps are sufficient or if it needs polyporpylene.

Filtering is mainly an option if the instruments to measure or use the reference are especially sensitive to a certain not too low frequency band. The main slow part would be someting like the DMM auto zero cycle with maybe 2-4 readings of 200 ms each. So this would be frequencies of some 1-4 Hz.
Instead of the effort for filtering a lower noise reference could be the easier way.

[guenthert]

[..]
For filtering I would not consider electrolytic caps practical, because of the temperature effect and leakage that can change with time and temperature. The DA will also requite extremely long settling - maybe not as long as the LM399, but it would start all over after power off.
[..]

      Is DA really an issue, if one uses two capacitors in series and charges the mid point from the output of the zener using a 2nd resistor (as described in AoE3)?

[Cerebus]

That little trick is designed to deal with leakage currents, not dielectric adsorption.

[guenthert]

That little trick is designed to deal with leakage currents, not dielectric adsorption.

      Well, yes, but what is the ill-effect of DA, if not voltage drop over the resistor due to the charging current?

[Kleinstein]

The trick with 2 capacitors would also suppress the effect of DA, at least to a large part. There is still the DA for the lower capacitor (that still sees the voltage). Because of the usually smaller resistor there, the extra current from DA would have less effect, but could still be an issue for some time. Still more like minutes or hours instead of weeks. Some of the temperature effect can still be there, even with zero voltage at the capacitor.

[iMo]

The substrate polarization - is the wiring with H- and Anode on the same potential (ground) comparable (in relation to achievable Vref parameters) with the wiring where H- is somewhere at -15V, for example?

[eplpwr]

Where can I find the ADR1399 datasheet?

It is now released on the AD website, complete with datasheet Rev 0.

For those of us that have a VAT# the ADR1399KHZ is orderable at US$ 12.88 ea, MOQ 18 pcs. I also saw the "Request samples" button was available, though didn't try it.

[maat]

I also saw the "Request samples" button was available, though didn't try it.

I works  . Delivery date pending...

[Cerebus]

I also saw the "Request samples" button was available, though didn't try it.

I works  . Delivery date pending...

Yup, it does.  I've got "Backorder Allow 48 hours to be scheduled" as the current status.

[magic]

I collected several MAC199 (LM399 clone in TO18 case) with broken heater. I've tried to utilize the parasitic diode for chip temperature measurement.

Basically it works, with say 80uA fw current the voltage at the heater's ground changes like (aprox)

31C (fingers)  620mV
26C (amb)     637mV
IPA large drop  677mV

There is an issue with it, however - while the 80uA current is injected into the heater's gnd (parasitic diode anode) the Vref (standard 10V Vref like the above) increases by +3000ppm. That is way more than a drop on the bond wire - thus it messes up something on the chip, it seems..

Bringing the substrate above circuit ground may forward bias parasitic diodes from the substrate to circuit elements like NPN collectors.
Try biasing the substrate normally and using the diode going from the substrate to H+ instead. (If the heater failure is not due to H+ being open circuit).

[r6502]

Hello all,

I've had supid idea:

I do not own a 7.5 or 8.5 digit multimeter, only 2 PREMAs and one HP 34401a, so all 6.5 digit types. These devices are using the same reference as my reference board that I'd like to characterize.

My thought was based on better thermal stability of LTZ1000  based reference design. I thought I will replace the LM399 in on of my PREMAs and put a LTZ1000 based reference in. The 7.2V could be direct used in the PREMA. Due to the higher stability of the LTZ100 I will now be able to see the drift of the LM399 based 10V reference, if the drift is > 10µV.

What do you think about this Idea?

Edit: may be I will use one of the new AD ADR1399 references for the 10V reference board, when I get some ...

[Echo88]

Improve the stupid idea: Use the LTZ1000 as a standalone stable reference and take a 34401A to measure the voltage difference between a DUT like the LM399 and the LTZ1000 in opposition mode (ref-minus connected to ref-minus).
Dont really know if a LTZ1000-modded 34401A would show better stability or if its limited fundamentally by its ADC/Frontend.

[branadic]

I totally agree with Echo88, at least the Prema DMMs are not limited by the noise of the reference, but from the ADC. Using them in the 100 mV range to measure the difference between two references makes much more sense.

-branadic-

[Andreas]

Hello,

most LM399 references that I have have less than 2 ppm voltage change near room temperature over a 30 deg C span (10-40 deg C)
2ppm/30 deg C = 67 ppb/K.
A LTZ1000 is usually not better if it is not trimmed for T.C.
So you have a hen and egg problem.

Attached a measurement between 2 LM399 (one at room temperature and one in my car cooler box)

With best regards

Andreas