JCan kits and parts
allows you to measure the Johnson (thermal) noise of a resistor in a home or school lab.
It appears from some comments we recently received recently that some people (none of whom actually built or tested the experiment) have lost the context of this experiment. 
The original JCan Article, Reprinted by permission of T&L Publications, Inc. (Many thanks!)
Building the experiment is fun and interesting and creates a unique home lab instrument. Use our MS Excel calibration worksheet to convert your results directly into standard noise readings in nV/rt Hz . Instructions and project details can be found in the July, 2007 issue of Nuts & Volts Magazine . More information on resistor noise measurements can be found on our JCan References page. More information on measuring the noise bandwidth of a filter can be found in our nBW Technote. To perform the JCan experiment and use the JCan instrument, you will need an AC voltmeter that can measure between 1 and 40 millivolts AC. Most modern bench meters (DMMs) have a suitable AC voltage range. 
Overview: Once calibrated (two numbers entered into the Calibration Worksheet), plug a resistor into the RUT jacks (with the power jacks open), turn on the power, close the lid, and measure the output amplitude in mV AC. Plug that number into the worksheet and get a thermal noise value in nV/rt Hz for that resistor. The blue worksheet curve represents the theoretical Johnson noise curve at no current. The green curve shows how your measured data falls on the curve. Finally, use two like valued resistors, one in the RUT test jacks and one in the resistor power test jacks, to see how the noise changes for different types of resistors under power (e.g. metal film as compared to carbon composition). Use the parallel value of the two Rs and plot a corresponding noise point above the Johnson noise curve. If you have an audio spectrum analyzer or a FFT capability, observe how the 1/f noise for a pair of carbon composition resistors is much higher than for other types such as metal film resistors.
This page changes often. Please hit "re-fresh" or "F5" on your browser to be sure to be looking at the most current page.
There will be a limited number of kits, parts packs, and assembled JCans available in this first run as we judge interest in the JCan product line.
JCan Kit (quantities are limited, but still in stock): $79.95 A kit of all parts to build the JCan shown in the article. The kit includes the JCan PCB, all semiconductors, resistors, capacitors, pins and sockets, an undrilled 1/2 gallon can and a resistor "cal pack" for calibration experiments: (two 9V batteries2, BNC cable, AC voltmeter not included) (For European Union (EU) buyers, the kit is not RoHS compliant) Please add the cost for shipping 1 pound 1 ounce, in the U.S. from zip 13492 to your zip (Priority Mail). Use "package", online price, Priority Mail. We tried to keep the package under a pound for standard rates, but there is just too much stuff in the can :)
Please see our Ordering page for payment options. We accept university purchase orders.
Assembled and tested JCan: No longer available $289.95 with calibration data including a bandpass plot and completed JCan Worksheet with measured nBW, noise floor, gain, spare front end JFets, spare gold posts, and a printed copy of our JCan article (two 9V batteries  , BNC cable, AC voltmeter not included) . Please add the cost for shipping for 3 pounds (16" x 10" x 8") in the U.S. from zip 13492 to your zip (Priority Mail) Use "package", online price, Priority Mail. (Allow up to one to two week delivery times from day of order. It takes about 6-8 hours to assemble and fully test a JCan with calibration data).
JCan PCB $8.95
Gold Jack Pack $5.95 (four gold slit jacks, 9 gold MillMax pin jacks, 2 IC sockets gold insert/tin pins)
Resistor Cal Pack $2.49 (fifteen 1% metal film)
For a few small packs add $2.33 for First Class shipping in a small box in the U.S. (or, $5.35 for Priority Mail for quicker delivery).
All checks and money orders should be made payable to "Joseph M. Geller"; please see our ordering page for more information.
Sales to EU electronics hobbyists and Amateur Scientists: Some parts of the JCan kit and assembled unit are not RoHS compliant. While we cannot actively market a non-compliant item to the EU, we have written advice from the RoHS UK authority, that it is permissible for EU electronics hobbyists and Amateur Scientists to purchase one unit for their own hobby use so long as it will not be used for their commericial gain by re-sale in the EU. EU buyers should contact us please for shipping prices. Modest cost shipping cost to the EU can be accomplished by using International First Class Mail.
Prices reflect our cost in ordering parts from various distributors and the time to make up the packs and kits. It is very time consuming to prepare many of the packs. In doing cost comparisons, remember to account for the shipping and handling charges charged by each distributor. We charge only the actual USPS postage (we do our best anyway, occassionally an estimate is a little high or low), with no charge for a sturdy shipping box (from Associated Bag or Uline) and no additional handling charges. Our very modest profit mostly goes towards operating expenses, including this website, shipping supplies and other overhead costs, and development of our next project!
1 Most Agilent, Fluke, and hp digital bench multimeters (either true RMS or average responding) as well as analog AC voltmeters such as the hp 400 series easily cover this range. Suitable analog or digital voltmeters dating from about 1967 to 2007 are widely available new and surplus at prices ranging from $5 to $150 for an older tube or transistor analog meter typically needing some restoration, $50 to $250 for a Fluke or hp 4 1/2 to 6 1/2 digit digital bench meter from the 1980's, $400 to $800 for a used 6 1/2 digit hp or Agilent 34401A from the 1990s, to $1,300 for a new Agilent 34410A. Some hand held analog and digital multimeters may not have an ACV scale sensitive down to 1 mV AC.
2 The two 9V batteries are not included because they are relatively heavy and might come loose during shipping causing damage to other parts.
3 By focusing our measurements at zero resistor current (as nearly as practical) our goal was to make a Johnson noise measurement across a range of resistances, -at very low cost- with a -relatively simple circuit-. (None of these observers built the experiment or tested any part of it either in simulation or in hardware).
Once we showed that you cannot simply place a resistor on the input terminals of a standard voltmeter (even a very high-end bench DMM) to measure resistor Johnson noise (which is how the project got going), the goal was to find a low cost relatively simple way to measure the Johnson noise of a resistor. Although we do include a demonstration which contrasts a carbon composition resistor under power to a metal film resistor under power, our goal was not to rigorously quantify the differences between types of resistors under operating conditions, a job well beyond the scope of this experiment.
By definition, the Johnson noise is just the theoretical thermal noise. Johnson noise in and of itself, does not include the other components of total resistor noise in operating conditions.
Of Course, when a resistor is powered, different types of resistors can and often do exhibit widely different noise characteristics. Total resistor noise at operating current includes the Johnson noise plus many other noise mechanisms. When we added some bias current at the end of the experiment, the point was simply to show that such differences can be observed with a relatively simple experiment. We noted that when an audio range spectrum analyzer or scope with FFT function is available, the "powered resistor" spectrum is interesting to look at. However, we do not rigorously quantify those relative differences with the JCan experiment. The JCan experiment was not intended to replace commercial instrumentation such as the classic QuanTech resistor noise analyzers of the '80s (e.g., the classic 315C which probably cost several thousand dollars, at least in today's dollars) which quantified total resistor noise under power, or their modern equivalents.
We hope, perhaps in part because of the JCan experiment, that more students and hobbyists are thinking about Johnson noise.
On JCan Input Capacitance: As explained in the JCan article, in order to achieve valid thermal noise measurements with minimal error (on the order of 1%) beyond kilo ohm values, the preamplifier (our JCan amplifier) must have an ultra-low input capacitance. Otherwise, an in-band portion of the actual thermal noise voltage (within the JCan bandwidth) would be attenuated at the input posts due to the first order RC roll-off of the R under test in parallel with the input capacitance. Beyond inference of input c, made by comparing measured results to the to the theoretical Johnson noise curves and observing the increasing deviation (high end roll-off), we had tested input c by using a relatively high valued input resistor where we tested for the -3dB point with increasing test frequency. More recently I came across Jefferts, Steven R.; Walls, A very low-noise FET input amplifier, F. L., Review of Scientific Instruments (ISSN 0034-6748), vol. 60, June 1989, p. 1194-1196. Jefferts and Walls measured the input c of their low noise amplifier by recording the drop in output voltage with the addition of an input series capacitance. Using a 100 pf capacitor (measured 106 pf at 10 kHz with an Agilent U1733C LCR meter), I tested one of our sample JCans on the shelf with a 10 kHz, 352 uV rms input voltage (Agilent 33210A with a Kay 839 step attenuator). With no series input c, the 352 uV input value was chosen to set the output voltage to 1.0 V (hp 400GL AC voltmeter input voltage monitor, hp 3400A AC voltmeter output monitor). When the 106 pf capacitance was inserted between the signal and the input terminals, the output voltage fell to 0.94 V. This sample measurement of the drop in output voltage (.94 ratio) for the addition of the 106 pf capacitor, corresponds to an input c of about 7 pF.
One commentor on the JCan project noted that in his reading of the schematic diagram, the input capacitance, in his opinion, had to be at least >40 pF because of the Miller effect.
This is not the case because of the virtual AC common provided by the OpAmp in our composite BF244 - LT1028 input circuit. That is, the small signal RL is the 1 kohm DC resistor which sets the DC bias point in parallel with the 1 uF film capacitor to virtual common. Thus, the JFET common source stage Av taken alone is <1. Since the input capacitance due to the Miller effect is proportional to Av times Cgd (JFET gate to drain capacitance), the Miller contribution is negligable. Therefore our BF244A stage taken alone (out of context from the BF244A-LT1028 composite front-end) serves as an ultra low noise, low capacitance impedance converter having a negative gain (typically around -10 dB). The composite stage, however, has a positive gain, typically around +40 dB). A frequently asked question is, why didn't you just use the LT1028 as the input stage. The answer is becuase the input capacitance is too high to achieve good measurements (on the order of 1%) for higher input Rs.
Note: the correction factor from average responding to rms for noise is 1.13. The "2" superscript in the article is a reference to the Horowitz text footnote, not "squared".
All prices are subject to change without notice, all sales are subject to availability of parts and kits which are in limited supply.
COPYRIGHT © 2007, 2012 JOSEPH M. GELLER