Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x

[branadic]

Found almost zero t.c. current for that specimen at about 3.29mA which equals a 1.2k resistor on 10V.

-branadic-

[branadic]

OT: what would happen if you feed the diode with a simple 2xPNP 5mA current mirror?

Why should you? My goal is to use a circuit as shown in the first post of this thread, thus the boosted zener ouput voltage (10V on the opamp output) is feeding the zener itself through the current setting resistor. This is why I try to find to optimal current setting resistor while powering everything from a low noise 10V source for now. Next step is to calculate the optimal resistor values of the boost stage to get 10V.

-branadic-

[branadic]

Measured a little bit more at the italian sample and found, that this device shows excess popcorn noise. Attached are measurements of multiple temperature runs with 1k2 (3.29mA) and 1k247 (~3.16mA) as well as a 10h measurement at 3.29mA.
Need to investigate that a little deeper.

-branadic-

[Kleinstein]

The popcorn noise looks really nasty / high amplitude. Give that these references are not that expensive, I would consider this one defective.
If at all it might be a candidate to investigate / demonstrate popcorn noise.

[SZA263]

The noise seems too high in amplitude and too regular in frequency. The jumps seems to appear every '500' of your time scale, and seems to not appear during cooling phase.
IMHO is a residual switching noise from the heater. Try to rerun a cycle with heater switched off, disregarding the absolute voltage variation, only to see if the noise is still present.

[branadic]

No, you can see the jumps in the temperature cycling measurements during cooling phase, which takes much longer then heating phase and includes a temperature lag between case temperature of the reference diode and the actual die temperature. The heater is simply turned on and heats up to ~38°C. I then wait a while for everything to temperature stabilize and I then turn of the heater. Because of the thermal mass and thermal capacity (aluminium rod) the setup slowly cools down. It's not a controlled cooling, but a cooling due to convection into free air. So there is no switching noise of the heater.

In the 10h measurement (last picture) the temperature was almost stable at about 38°C (heater was turned on for the whole time). So I don't think it is a heater switching, since the heater is not switching but regulating.

EDIT: I currently have a measurement at higher current (~5mA) with heater on (~38°C) running and up to now can't see any jumps, which also seems to prove that it is popcorn noise. The disadvantage is, that a better temperature stability is necessay, as this operating point is on the positive side of the t.c. curve with ~7ppm/K.

-branadic-

[branadic]

Measuring long enough at 5.1mA (770R) proves there is (almost?) no popcorn noise visible and disproves the theory that the noise is due to the heater.

-branadic-

[SvanGool]

If I look at the discussion here and further: https://www.eevblog.com/forum/metrology/ultra-low-noise-reference-2dw232-2dw233-2dw23x/msg1054431/#msg1054431, you are unfortunate because you are not measuring a device which has the zero TC and low noise point close to each other. I would quickly test some other devices that you have (hint  ).

[branadic]

No, I'm not unfortunate, it's simply scientifics everyday, sometimes you loose and sometimes others are winning
I repeated the 10h measurement at 3.29mA and again can see excess popcorn noise so it's neither the heater influence nor a measuring artefact, but fact.

So the only option is to use this device at 5mA and because of positive t.c. with a more stable oven. Work in progress...

Meanwhile I can again take a look at the samples I received from the netherlands

-branadic-

[Andreas]

Hello branadic,

according to data sheet the zero T.C. point is more at 65 deg C than at room temperature.
Perhaps worth a try.

with best regards

Andreas

[branadic]

Well, with respect to blackdog even shorting the resistor on the crystal heater would only give 50°C.

The oven
The first problem is the temperature of the QH40A, which is about 41C and too low for my application.
The components that I mount on it also generate heat and then it goes roughly at 35 to 36 degrees, I estimate.
That means that kidney can be arranged more properly.
So ... that is "Reverse Engeneration" and that is successful, the circuit is very simple.
An opamp and a bridge circuit with a PTC of 2K in one leg.
The manufacturer has probably already taken into account a different temperature.
After I had drawn the diagram and what had counted on it turned out pretty simpe.
There is a 560 Ohm resistor in a part of the bridge, if you short-circuit it he will cut to about 52C.
Exactly what I needed to have him arrange reasonably on a summer day.

So for 65°C it needs a different oven design. Where did you find 65°C? I had in mind that a t.c. value of 5ppm/k between 25°C and 75°C is given in the datasheet.

-branadic-

[Kleinstein]

Hello branadic,

according to data sheet the zero T.C. point is more at 65 deg C than at room temperature.
Perhaps worth a try.

with best regards

Andreas

The TC depends on temperature and current.  A relatively small change in current is sufficient to shift the temperature for zero first order TC. The curve in the DS is just an example curve.

[Andreas]

So for 65°C it needs a different oven design. Where did you find 65°C? I had in mind that a t.c. value of 5ppm/k between 25°C and 75°C is given in the datasheet.

https://www.eevblog.com/forum/metrology/ultra-low-noise-reference-2dw232-2dw233-2dw23x/?action=dlattach;attach=509624

See figure 1 (Typical curve)

ok its more 55 - 60 deg C for the flat top.

But it is a typical curve (so some reserve in temperature range cannot be bad).

with best regards

Andreas

[Kleinstein]

With changing the current one can adjust the linear TC and thus the Temperature where the zero crossing will occur.  With a just slightly different current one can usually (if the unit is OK)  get it at any point in the usual temperature range (e.g. -50 V to +150 C). There is nothing special about 55-60 C

So it is OK to first design the oven part for a suitable temperature and than adjust the TC via the current.  The oven temperature has to take into account the self heating that can be quite a bit if 20 mA are needed. 50-60 C could be a suitable temperature if high ambient is allowed for. With a moderate ambient it could be 45 C.

[branadic]

Even though it's not directly related to the topic I use this place to correct for the circuit of the QH40A heater here, as I've already linked to the circuit.

So it is OK to first design the oven part for a suitable temperature and than adjust the TC via the current.  The oven temperature has to take into account the self heating that can be quite a bit if 20 mA are needed. 50-60 C could be a suitable temperature if high ambient is allowed for. With a moderate ambient it could be 45 C.

Well, that's only half of the truth... for a portable reference you want low current, but also zero t.c. so you need to find a sample with zero t.c. at idealy room temperature on lowest current. Furthermore you have to watch for popcorn noise at this low current and prove that it's low noise. At the end it's a trade-off between temperature, current and noise.

-branadic-

[branadic]

Did some measurements on the 5x 2DW232 samples from the netherland this weekend...

-branadic-

[iMo]

Above diodes: that means 50-60ppm/C in an average?

[Kleinstein]

The later diodes (no 2-5) seem to need a considerably lower current for low TC.
So these samples seem to be very much on the low current side opposed to the samples from the other source.

It is quite annoying having so much scattering. I thing they should do a little more binning / sorting in china, so one would know wether one gets a diode good for operation a 1 mA in series mode for a low power reference or one good for 5 mA in series mode if power is not an issue or one good for 1-2 mA in Zener only mode for a 5.2 V reference (to divide down to 5 V with little errors from the resistors). All 3 cases can be interesting, but different applications and normally different part numbers.

[GigaJoe]

this one , made in the 17th century, I think,   covered with patina.
for over 100 years,  it should be pretty stable

[branadic]

Measured a lil' bit on a first sample 2DW233 from the netherlands. With 10ppm/K @ 7.5mA it's still not meating my understanding of t.c. given in the datasheet. So maybe Andreas is right and the spec means, if you draw a line between 25°C point and 75°C point this line has a slope of 5ppm/K. Such a value would be useless as it ignores everything in between, so I take it as a marketing gag.
However, I can find an operating point at a current of 4.3mA so I can operate it heated with my crystal heater. Need to look for noise.

Still have another 4 samples for testing.

-branadic-

[Kleinstein]

The scattering in parameters seem to be rather large. So compare the 2DW232 more too an 1N821 than 1N829.  If the current for zero TC is still in a good range (like 3-10 mA) this is not such a big problem. To get a low TC one would need an individually matched current anyway. A small search range helps a little, but not that much.

The bad ones are those that need more than 10 mA in series mode, which would be quite a lot of self heating. These might still be of some use in a different mode (e.g. zener only or 2 zeners + 1 diode).

[GigaJoe]

Well, i have 233
and connecting in parallel 2 internal diode, simply. TC is "U" shape ; if temp UP  and low current  V down,  high current V up.  to  find a neutral spot is a time and patience,  my record approx. +0.15ppm/C. 

so from that above, that has a month in 120-150C and current 25ma im building 6 in parallel,  in assumption that some has negative and some positive TC,  due to current , and able to self compensate.

btw feeding by precise 10V a low TC spot offer 1023 Ohm resistor. after month it 880 ohm,  so parameters  i assumed drifted significantly.
233 2 diodes in parallel approx 4ma. for low TC, to adjust low TC spot i use 20ohm multi-turn trimmer.   using an owen even more fun as possible to compensate even  current resistors drift, due to positive or negative diode TC.  just adjust the current
 
"after month" - I mean 30 days, under 140C and 25ma.

they oxidize significantly,  even legs lost some coating and to tinned it back i have to use some crazy flux.

[IconicPCB]

Giga Joe,

What is the spread of currents ( resistors at nominally 10V )?

[GigaJoe]

Uh ... I'm actually just started building the  board,  so all this measurements and so ..  And tell precisely later what happens ... But as I can recall spread from 4.1ma to 4.6 ma per diode in parallel I think i played with 4-5 of them. The biggest trouble to use it , was an internal resistance,  60-80 ohm,  means even slightest current change and everything changed. Like in comparison LM399 0.1 ohm but yes it not  a pure zener. 

[Kleinstein]

The higher differential resistance is a downside of using it at a lower current.  60-80 Ohms is really quite high for a current around 4 mA.
The LM399 is not a fair comparison here, as it includes an internal amplifier and current source. One should more compare it to zener references like the 1N825 which gives <10-15 Ohms  at 7.5 mA.  So yes it needs a reasonable stable resistor.

Operating in the 4-5 mA range is nice, as it limits self heating to a reasonable level. I don't think one would use several in parallel of these, as the noise level should already be low enough for most practical applications. More like using 2 in series (or a bridge circuit) to get a reference of around 10.5 V and than divide down to 10.