The New Test Signal

Phi Spectral Contamination
My new test signal is as much a concept as it is a specific set of frequencies put together in a certain way. Traditional test signals either use a single sine wave tone or at most, a pair of tones, to explore simple harmonic distortion, and simple IM distortion. Music is composed of many tones, all happening at once, and with varying relationships. - - Multitone test signals use more than one single sine wave, more than just a pair of sine waves, in order to more thoroughly excercise the system. Before I came up with the new signal, the existing multitone signals as used, and as available via major hardware and software had some common faults. One of these was the use of regularly spaced tones, with spacing intervals that tended to be the same.

For instance, one of the popular test signals had tones at 100, 200, 300, 400, 500, 700, etc and at other frequencies as well. The problem with this is that the spacing for a lot of the frequencies is the same: 100 hz. This means that IM distortion generated from these particular frequencies would tend to have as one of the common frequencies a product at 100 Hz (200 -100, 300 - 200, 400- 300, etc.). Not only would all of these tones tend to have a similar distortion product frequency, but it would be impossible to even determine the composite level of these particular distortions, as the 100 Hz tone present in the test signal would overwhelm them. In effect, the measurement would be blind to these particular distortions.

I think that it is easy to see that there are many possible combinations that will end up covered up by the primary test signal tones, such as distortion products at 200 Hz, etc. There is also the fact that many harmonics will be covered up too. The harmonics of the 100 Hz tone at 200, 300, 400 and 500 Hz will all be impossible to detect or measure.
One of the current multitone test signals has tones spaced every one/third octave, which guarantees a lot of cover up of IM and harmonic distortions with this particular test signal. Not only that, it has 31 tones, which means that the overall level of each of the tones has to be kept low enough ot avoid clipping the system, whether it is the analyzer, the device under test, or whatever would be stressed. For 31 signals, this is about 30 dB down from 0 dB nominal.
After checking out what was out there, I realized that if you make all the tones fall at intervals where they would not have the same spacing ratio's, and would not have the same frequency intervals, this would avoid the cover-up of the distortion products. This idea in and of itself turned out to not be new, but after checking around, I determined that no one had come up with a consistent method to try and assure that as few of the tones, and all of their distortion products up to higher orders, would not interfere with one another, and not be covered up by the original tones. If you select frequencies at random, there could always be a bad combination, and it would not be easy to predict.

NEW! Link to paper on the Phi Spectral Test Signals

All applicable copyright laws apply, all rights reserved, except transmission by USENET and like facilities granted.  Any use or inclusion in print or other media are specifically prohibited.  The informational content is not warrantied in any way or form, and any use of said content are at the reader's own risk, the author shall not be held responsible in any way for any damages or injuries arising from the content of this web site.  Common safety practices are encouraged at all times.  Do not fold, spindle or mutilate.