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

[Kleinstein]

The noise part looks OK. It is however hard to say how much of the noise is from the DUT. Most of the noise could well be from the meter itself. It still shows that this ref. sample has no excessive noise.

The dift is some 20 ppm over nearly half a year. With the limited number of points it is hard to tell if the drift is really slowing down and the voltage settling. Another point is that the temperature may have changed between the measurements. So part of the drift could be a temperature effect. Finally the 6.x to 10 V gain step could also contribute to drift.

[iMo]

No idea what the temperature there is like (meter or ambient or reference?)..

[lichen867]

the root cause might be effect of stress. mechanical stress will be recorded by the chip，thus you will see different voltage even at same temperature.

[ZhuraYuk]

So I bought 10 of 2DW232 few years ago. Today when my low noise amplifier is ready (clone of LFLNA-80) it is time to measure if these are worth of my time.
So amplifier noise specs are following - 0.1 -10 Hz 17nV RMS 120nV PtP
Lowest noise level was observed at 4 mA current for any 2DW232  I tried. It is about 110nV RMS 620nV PtP. At 23mA like people are posting here the noise is few microvolts, very high.
Connecting 5 2DW232  in parallel gives around 58nV RMS and 350nV PtP noise levels  (10 Volts outputs, with op amp, two transistors and 5 ppm resistors in divider circuit)
Conclusion - if you need low noise levels this is very good solution, similar to ADR1001.

[Birb]

Out of boredom here is another attempt at creating a (scuffed) 10V reference from the 2DW23X
This one uses an LM35 and a very simple proportional feedback circuit to achieve <0.1C of stability. Temperature coefficient is ~10ppm/K, most likely due to the resistors chosen being typical pullup resistors. Surprisingly good for randomly selected devices.
Newer devices seem to be 6.4V to 6.5V typical, so an equivalent 6k:11k gives a very good ~10V output.
I don't have an LNA to measure noise so I cannot attest to it yet, but measuring with respect to an LM399 reference gave ~7uV RMS during the most stable period.

There were random jumps which are probably temperature or other influences. It is probably not (very) insightful since the data was acquired with a ~4.5 digit meter and many random jumps are present, either due to temperature or other influences.

Unfortunately, the reference mysteriously died ~5 days after initial power on.

Nevertheless, this has shown that low quality components are capable of making an "ok" reference, so here is another design involving a differential amplifier on the diode voltage, and measuring the voltage difference to a voltage setpoint derived from the zener voltage. This effectively supplants the LM35, and replacing the components with slightly higher quality ones should improve performance by a lot. The circuit board measures 42.5mm*30mm so it can probably be doubly ovenized.

Is it worth it to replace the diff amp with a common inamp such as INA128/9?

Thanks.

[analityk]

100-120 nV rms is this same level that I've measured and i have also used lna of Euler Precision with 80 dB gain. I have one diode in my circuit. It is a bit old construction, you may follow this thread, maybe 4-5 years.