Im surprised to see that those small modules like the A23 use solderable pins for the voltage input instead of copper-threads. How are they able to maintain those 1nV/°C-drift spec with solder-bridges (like on the filter-pcb they offer for the modules)?
A solder joint is not that bad: The solder gap is usually only very small ( 100 µm range) and the temperature gradient is small if you take care. In addition there as solder joints for both inputs (GND and input) and some of the errors will compensate. It can take some care with the thermal layout of the wires (e.g. keep both wires close together) - for magnetic coupling it is a good idea to do this anyway.
To cause an error due to themal EMF it takes different materials and a temperature difference that changes. A constant temperature gradient (e.g. due to self heating of the amplifier) would only look like a small additional offset and not have an effect on those Allan deviation plots.
I did some further stability testing of an HP3458A. I added the results to a plot with the HP34420A and the EM A10.
One HP3458A had half the noise of another HP3458A on the 100mV and 1V scales. I have a third unit I will check sometime. The data in the plot is the unit with higher noise.
The top line is the stability on the 1V scale with auto zero turned off. The line does not continue up forever. It levels off before 2uV. This will depend on the meter and ambient conditions.
Reference : "Unspecified Low Frequency Noise in Chopper Op Amps"
Sensors and Transducers Journal
The low-frequency input voltage-noise of precision op-amps may be impaired by external components
Does anyone have data on the modules (first from Philbrick and then from ADI) that used varactor diodes in the front end, energized through a transformer? I wonder how they compare in the 1/f range with solid-state choppers. Usually, people concentrate on the very low input current (+/- 10 pA for the Philbrick and +/- 10 fA for the later ADI) and huge input common-mode voltage (+/- 200 V). These amplifiers are very slow, with a 75 kHz unity gain frequency for the Philbrick. ADI's website has a link for the PDF datasheet from "obsolete" components, but it sends one to an announcement of brand-new parts. From an ancient ADI article, it looks like the 301 module had an input voltage noise of 70 nV/rt Hz and an input impedance of 1010 ohms at 25 C, so its voltage performance may not be as impressive as the current and common-mode performances.
I have some copies of data-book pages for the ADI310 and 311 models. The voltage noise is really not impressive: 10 µV_RMS for 0-1 Hz and 30 or 10 µV/K of drift. The varactor bridges use a kind of chopping / modulation, but the varactors already have a TC by themself (some -2 mV/K similar to a normal PN junction).
I've noticed some messages about problems when using GPIB on Keithley 181. Maybe this will help. Figured it out by looking at some source code by TiN.
1. You have to switch the meter to addressable mode (not talk-only) (on the back).
2. You have to select address (it's binary!)
3. You need to end every command with `X` (means: eXecute)
Sample code (Python, `interface` is a PyVisa Instrument):
interface.write('B1X') # 6 digits
interface.write('P0X') # filter off
interface.write('D0X') # damping off
interface.write('R1X') # 2mV range
interface.query('') # not needed really, read will be fine too
One quirk is that first response is weird, instead of `ZDCVXXX`, sometimes I just get `ZXXX`. But the rest works just fine, meter is responding to commands and will switch to desired modes.
I do not see a Keithley 2182 (non-A) version for comparison. Attached is an initial look at the performance with a custom LEMO short. Please note that I still need to verify that everything is working correctly.
The LEMO connector is shorted with a pure copper collet that I designed so that no solder is required. Insertion requires a good amount of force by the fingers to guarantee good contact. It is then insulated with tape before being assembled to prevent the inputs from being connected to the outer shield. Despite the 'significant' thermal mass of the collet, the settling time is very short, about 1 minute or so on first impression, when first inserted.
Based on the following digitalized information from the 2004 Keithley Catalog, my 2182 (non-A) version seems to work ideally.
Based on the following digitalized information from the 2004 Keithley Catalog, my 2182 (non-A) version seems to work ideally.
* Hilariously, the Specifications for the 2181A claim "lower measurement noise" compared to the original but copy the same data as the original to demonstrate the performance. -- Keithley got lazy.