This post is one of three related to the work done to measure resistor excess noise. A summary writeup and three other documents with experimental detail are provided to those interested in performing their own experiments. There is still ongoing work to test more parts, the revision currently posted is Rev 3.0
While this information is presented somewhat in the style of a research paper and its backing data, it is better thought of as a set of cleaned up (electronic) lab notes kept over the course of doing the work. While the intent was always to share the results among the ABW community, it was realized that the larger AES community might also be interested, so the notes were formalized a bit in to these documents.
It was also decided that the questions raised by this work might be of interest to student AES members looking to start their own research, so details where added to hopefully allow easier replication and extension.
This document is a short summary of the work done, and the best place to start.
By comparison the other documents contain a lot of detail and more than one dead end. A high number of plots also results in many pages.
This document looks at the problem, what others have done, and lays out the plans for the experiments.
The original plan included studying resistor distortion, as well as studying the effects of the excess noise and distortion in a typical op-amp circuit. As the noise project grew in size those other efforts were deferred to a future, separate effort(s).
A number of people reviewed the early work and provided excellent feedback and those credits appear in this document, along with a list of references.
Test fixture build and verification
This document looks at issue during the building and validation of the test fixture.
One of the key aspects of the design was to provide a “known reference” to use on the 2nd AP input when making measurements on the 1st AP input. The idea was that it was easy to make experimental mistakes, particularly when some of the parts had excess noise below the limit of the AP515. With a “known reference” always on the plot it was hoped that procedural errors would be easy to catch – and they were.
It turned out the parts chosen for the reference, 2.5 k ohm wirewound resistors, had a NI (Nosie Index) over 20 dB worse than all other other wirewound parts tested. Theoretically wirewound parts should be almost ideal parts, only metal foil parts might have lower NI. The cause of this outlier remains unexplained.
A change was made between the first tests (rev 1) and the second round of tests; the AP plot results were changed to the ASD (Amplitude Spectral Density) and 1/24 octave smoothing was used to make the graphs easier to read data directly from.
To enable easier comparison among parts all of the parts tested in rev 1 were retested following this newer workflow. The Rev 1 results were retained in the document in case later questions were raised. All of the Rev 2 retest results nominally agreed with the Rev 1 results, so after a certain point boredom set in and full retests were not done. Speaking of boredom, if you’ve read the other two posts and all of the papers listed here, you must be bored of the topic by now.