A few details on the schematic and block diagram
![[photo]](../images/research/electronics.jpg)
![[photo]](../images/research/summing_amp.jpg)
First, here are some slides from a seminar I gave on making electrical measurements. Toward the end, there is a description of how the dI/dV measurement works if you are not familiar.
The dc voltage is supplied by a Keithley 263 Calibrator/Source (K263), which can supply +/- 20V. The sample voltages we desire are in the range of 1V or less, and often on the scale of 1mV. In order to avoid discretization noise, we put out a relatively large voltage and attenuate it. The attenuation factor is selectable with a four-position switch: 20, 40, 400, 4000. For instance, we might source 4V from the K263, but attenuate it by a factor 4000 to supply 1mV to the sample and its wiring.
The ac voltage is supplied by the output of the Stanford 830 (SR830) lock-in amplifier, which has a range of 0-5V. The ac voltage is used as a modulation for measuring dI/dV, for example, and therefore must be small compared to the dc voltage. In extreme cases, it might be only 20uV peak-to-peak. As with the dc voltage, we apply a fairly large signal from the SR830, and attenuate it by a large factor, selectable in 5 ranges from 1 to 10k. This ac modulation represents the dV part of the dI/dV measurement. In order to have a constant resolution measurement, its magnitude must remain constant. However, since we typically measure non-linear elements such as tunnel junctions, it is not sufficient to simply set this voltage once for a given measurement. Software feedback (a simple PID loop) is employed to constantly maintain the ac modulation on the sample of interest to within 10% or better of the desired value (more details below).
A dI/dV or IETS measurement involves sourcing the dc sample voltage with a small ac modulation voltage superimposed on it, which means we must add the two source voltages. The attenuation for both ac and dc voltages is performed by a home-made amplifier, which also serves to buffer the two signals from each other. They are then added in a 1:1 ratio with a home-made summing amplifier.
Measuring the sample voltage is accomplished with two independent leads to the sample (for a true four-point measurement), which are fed into a battery-powered EG&G 113 wideband amplifier (gain 10-10k, selectable lo- and hi-pass filters). The dc portion of the signal is measured by a HP 3478A digital voltmeter to measure the dc voltage on the sample, while the ac voltage on the sample is measured by the same SR830 lock-in amplifier that supplies the ac modulation. Since the sample resistance is not always large compared to the wiring (not to mention the fact that the amplifier gain might not be precisely constant) it is necessary to separately measure the actual ac voltage on the sample. The ac voltage measured on the sample is compared with the desired ac modulation given by the user, and a simple software PID loop adjusts the lock-in amplifier output to maintain the actual ac voltage on the sample to with 10% of the desired value.
Measuring the sample current is accomplished by a Keithley 428 (K428) current pre-amplifier in series with the sample of interest. The current pre-amplifier appears (essentially) as a zero-ohm load to the sources, and outputs a voltage signal proportional to the current through it. It has a selectable gain of 10k-100G V/A, so measuring pA currents is possible on a good day. This signal is fed directly into a HP 3458A digital voltmeter to measure the dc current. The ac current signal (the dI part of dI/dV) is fed through a battery-powered EG&G 113 wideband amplifier (gain 10-10k, selectable lo- and hi-pass filters) and then to a Stanford Research 830 lock-in amplifier to measure the dI signal. This lock-in amplifier is synchronized with the ac source from the first lock-in at f (for dI/dV) or 2f (for IETS).
The EG&G amplifiers are necessary for two main reasons. First, the ac voltage and current signals can be quite small, and amplification can be necessary. Second, the input impedance of the lock-in amplifiers is only 10MOhm, so these high-input-impedance amplifiers act as buffers when measuring high resistance samples. The current pre-amplifier is preferable over a simple resistive shunt to reduce noise: having a small resistance in series for measuring current leads to small and noisy signals, while having a large resistance in series leads increased Johnson-Nyquist noise.
A few observations: it is hard to beat the old EG&G amplifiers, they are worth their weight in gold! Don’t thrown any of that vintage stuff away! The Stanford lock-in amplifiers are also fantastic, and very easy to use and program. Ditto for the Keithley 428, though it is overkill for many applications (the *lowest* gain setting is 1k V/A, so anything over a few mA and you’ll need to remove it). The summing/attenuation amplifier works beautifully, and the schematic linked above and implementation are thanks to Danny Whitcomb, who spent many hours pondering my strange request.