The forward and reverse characteristics match each other at very low
voltages. After all, the diodes characteristic curves are continuous through the origin (zero volts, zero current), and until the depletion zone width is maximized by reverse voltage biasing, or minimized by forward voltage biasing, the diode characteristic is essentially exponentially resistive with a temperature coefficient attached. Below is a typical diode characteristic curve plotted on
linear axes.
What is interesting to me is the deviation from linear (resistive) response that appears to begin very early, at less than 10 mV forward OR reverse bias. Maybe this helps explain how solid-state "crystal" diodes can rectify (detect) microvolt RF signals, because, if I understand the diode equation correctly, there
never is a linear response in either the forward-biased region or the reverse-biased region. Although It
appears that there is from the plot of your data on a log-log graphic scale, I think perhaps this just establishes the nature of the exponential response. Whatever, the reverse saturation current Is is important, and apparently lower is better.
I think I will see if I can duplicate your results with my 24-bit resolution Mooshimeter, after I find and install a micro-SD card to collect sample data on.
Your curves look too smooth to have been influenced much by noise injected externally. There
might be a problem with thermal emfs if dissimilar metals are involved in the connections, and those junctions are not all at exactly the same temperature. Lots of gotchas can be involved with simultaneous microvolt and nanoampere measurements. I will try to post Mooshmeter measurements in the next few days.
And to answer your question: Yes, I have often wondered exactly what happens near the origin with diodes. Maybe now that we can both afford some spiffy instrumentation it is time to find out!
Google "low level measurements handbook Tektronix" for more information.
Or download a PDF copy from
here. Be careful NOT to sign up at one of the many monthly subscription download sites. The link I provided came from the
Tektronix-Keithley website and was provided after a registering on that site.
BTW, best to use Leclanché dry-cells as voltage sources to minimize noise in the source supply, but even then the noise varies depending on the manufacturer... but you already know that. This used to drive me crazy when trying to bias photoconductive infrared sensors, until a more "seasoned" technician informed me that not all dry cells had equally low noise characteristics. Don't remember which brand was "best" or whether it is even manufactured in this Century. Easy enough to check them for noise before using to measure anything important.
