本帖最后由 xDevs.com 于 2024-3-10 11:18 编辑
Some years back in 2019 I wanted to experiment with lowering noise on DC zener reference. Best commercial reference - LTZ1000A has ~1uV noise peak to peak 0.1 - 10 Hz. This is only ok for 7 digit, but not 8 digit.
Because we don't have lower noise and better stability DC reference without quantum standard - we don't have 9.5-digit DMM in 2024. But what is one combine multiple LTZ1000 together into averaged output? If we take 4 chip noise should reduce by SQRT(4) = 2x times.
So we should see 0.5 noise pk-pk. If we increase zener current from 5mA to 10mA we can further reduce noise. So I designed quad-zener reference board and here are some results and photos.
Later I also build same reference device but with new ADR1000 chips (500 nV noise pk-pk 5mA).
Each zener core cell is individual output, opamp, resistors, heaters. Only the power regulator input is same (LT3045 low noise +12V). This let me to use reference as 4 separate 7.1/6.6V outputs or combine with passive resistor network to averaged low-noise.
PCB is 6-layer RO4350 to reduce effects from humidity on SMT opamps (TI OPA2140) and SMT resistors/capacitors.
Assembly of all SMT parts done with reflow owen. LTZ chip and VHP resistor mount by hand with low temperature to avoid thermal stress.
One of ideas to try this low-noise reference module was to upgrade HP 3458A :
And measured noise with 4 x LTZ1000A chip, 10 mA current, using 80dB enjoydiy amplifier 0.1 - 10 Hz:
250 nV peak to peak @ 7.1V , pretty good :). xDevs.com QVR consume about 160mA @ +13V and 5mA @ -13V for power.
I also tested many more ADR1000 and LTZ1000 chips (125 pcs total). ADR1000 noise is very good, usually about 450-500 nV pk-pk.
Later in 2022 and 2024 I've built some more modules with ADR1000 to test long-term drift and noise too.
Three QVR module (4 ADR, 2 ADR and 1 ADR) are tested for long-term drift with nV-meter and DataProof 160A scanner. Drift compared to calibrated and stable Fluke 732A and LTZ1000A xDevs.com FX units.
ADR1000 show drift about -2 ppm/year when running at lower +50 °C oven. No special aging or preconditioning was done before assembly.
 
Later I built separate module with mixed LTZ1000ACH, LTZ1000CH and two ADR1000 different datecode.
Tested and adjusted for good temperature coefficient using "400kOhm" feedback resistor fine-tune:
And final block diagram..
Hope for good results in 2025...
Low thermal TBP3 posts were used for best 5-way connection interface to the QVR reference module.
GitHub with results and more details : https://github.com/tin-/adr1000/
|
发表于 2024-3-10 11:08:32
提升卡
置顶卡
变色卡
发表于 2024-3-11 17:12:34
