This scoring utility provides a special automated algorithm for producing a comparison score reflecting the true sound quality of a car audio system. This routine was developed by LinearX during August of 1996 to solve some of the present problems associated with existing car audio scoring formulas. There are two principal problems which this scoring system resolves: (1) the score will more accurately reflect the true sound quality of the system, and (2) there are no hard threshold ranges which cause scoring variability.

It is well known that car audio systems scored under existing methods can have a wide range of system frequency response, while all may be producing nearly identical scores. This is due to the 3dB band/pair minor deviation requirement. Many systems using these older scoring formulas could have response characteristics with substantial differences in level between the low and high frequency ends of the spectrum. While the actual difference between any band/pair may be less than 3dB, producing a relatively high score, the overall system frequency response is extremely poor. In spite of these systems producing a seemingly high score, there was very little correlation between the score and actual real listening quality.

The second problem has to do with the repeatability of the scoring process itself. In the old scoring formulas, if a band/pair is less than 3dB no point is deducted, and a difference of greater than 3dB causes a whole point deduction. Any particular measurement of a given system can produce a number of band/pair differences which are right on the edge of the 3dB threshold. For example, the difference between a given band/pair may be 2.9dB on one measurement. During another measurement of the same system, the difference on the same band/pair may be 3.1dB. Even though the measurement changed by only 0.2dB, the score changes by one whole point. It is very possible that there may be several band/pairs in a system which are right on the edge of the 3dB threshold. This can cause the score to change by several points during repeated measurements due to the various band/pairs toggling at the 3dB thresholds.

The TSQ formula solves both of these serious problems, but also goes further. Many of the old scoring formulas use bands from 25Hz to 20kHz in the scoring computation. The pink noise source generally used for scoring is generated from a CD. The CD process cannot record frequencies above 20kHz. A bandpass filter centered at 20kHz has one-half of its effective pass band above 20kHz. Any measurement using pink noise from a CD will show about a 2dB loss in the 20kHz band, solely from the CD-DSP processing itself. For this reason the TSQ scoring formula does not use the 20kHz band.

Many of the competitors utilize 1/3 octave graphic equalizers with 28 bands, ranging from 31.5Hz to 16kHz. To match this equalization range the TSQ formula also only uses low frequency bands down to 31.5Hz.

The TSQ scoring formula is straight forward and easy to compute. Essentially the response of the system is scored based on its relative flatness, or reproduction quality. The scoring computation steps are:

(1) Find the median level. Add the dB levels of the 28 bands from 31.5Hz to 16kHz, and divide by 28.

(2) Find the total deviation. Add the 28 absolute value differences between each band's level and the median level.

(3) Find the total score. Score is equal to 40 minus the total deviation divided by 6. Round to whole number.

For the maximum SPL test, simply subtract 100 from the maximum SPL reached, and round to the nearest whole number.

For user information purposes, the computed median level is displayed, along with a list of the bands which deviate by more than 3dB, 6dB, 9dB, or 12dB from the median level.